smlflg's picture
Initial public upload from Projekte/PromptBoost (part 5)
638c806 verified
Raw History Blame Contribute Delete
2 kB
{
"id": "codex",
"label": "Codex",
"summary": "Outcome-first coding agent with local context inspection, scoped edits, and explicit verification.",
"section_style": "concise_numbered",
"evidence_level": "documented",
"claim_notes": [
"Documentation supports Codex as a coding agent; exact failure modes remain adapter hypotheses until eval-backed.",
"Use this policy for local repo work, patch review, tests, and bounded implementation tasks."
],
"sources": [
{
"title": "OpenAI Codex",
"url": "https://developers.openai.com/codex/",
"publisher": "OpenAI",
"accessed_at": "2026-04-21"
},
{
"title": "OpenAI Prompt Engineering",
"url": "https://developers.openai.com/api/docs/guides/prompt-engineering",
"publisher": "OpenAI",
"accessed_at": "2026-04-21"
}
],
"eval_run_ids": [],
"capabilities": {
"tools": true,
"filesystem": true,
"subagents": true,
"browser": false,
"approval_model": "workspace and command permissions",
"write_surface": "repo files and local artifacts"
},
"failure_modes": [
"Stopping after analysis without verification",
"Touching unrelated files",
"Reporting checks too vaguely"
],
"rules": [
"Start from the requested outcome and preserve the user's task intent.",
"Name concrete deliverables and quality gates.",
"Inspect relevant local context before changing code or claims.",
"Run available tests or checks and report what was verified.",
"Avoid unrelated refactors and unnecessary process detail."
],
"verification_gates": [
"Task intent preserved",
"Deliverables complete",
"Relevant checks run or explicitly unavailable",
"No unsupported task facts added"
],
"allowed_terms": [
"outcome",
"deliverable",
"quality",
"gate",
"test",
"check",
"verify",
"verification",
"local",
"context",
"refactor",
"implementation",
"codex"
]
}