diff --git a/LICENSE b/LICENSE
new file mode 100644
index 0000000000000000000000000000000000000000..d8fddab504f968b514eefe28ae5186b03b51f2aa
--- /dev/null
+++ b/LICENSE
@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 I-Lang Research, iLang Inc.
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.
diff --git a/README.md b/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..c98c9adb8fb87b3a71405a5badd1cc4d0b83d09a
--- /dev/null
+++ b/README.md
@@ -0,0 +1,375 @@
+---
+license: mit
+tags:
+ - ai-coding
+ - i-lang
+ - ilang-protocol
+ - claude-code
+ - cursor
+ - developer-tools
+ - vibe-coding
+pretty_name: AutoCode - AI Coding Skills Protocol
+configs:
+ - config_name: default
+ data_files:
+ - split: train
+ path: data/*.jsonl
+---
+
+
+# ⚡ AutoCode
+
+
+
+### You say it. AutoCode ships it.
+
+[](https://opensource.org/licenses/MIT)
+[](https://github.com/ilang-ai/autocode/releases)
+[](#-11-skills-7-stages)
+[](#-install)
+[](#-install)
+[](https://github.com/ilang-ai/ilang-spec)
+[](https://ilang.ai)
+[](https://doi.org/10.5281/zenodo.22865145)
+
+[🇨🇳 中文](docs/README.zh.md) · [🇯🇵 日本語](docs/README.ja.md) · [🇰🇷 한국어](docs/README.ko.md) · [🇪🇸 Español](docs/README.es.md) · [🇧🇷 Português](docs/README.pt.md) · [🇩🇪 Deutsch](docs/README.de.md) · [🇫🇷 Français](docs/README.fr.md) · [🇷🇺 Русский](docs/README.ru.md) · [🇸🇦 العربية](docs/README.ar.md) · [🇮🇳 हिन्दी](docs/README.hi.md)
+
+
+> **Every AI coding tool helps you write code. AutoCode helps you ship it.**
+>
+> You say what you want. AutoCode builds it, deploys it, and gives you a live URL.
+> You don't need to know programming. You just say yes or no.
+> When it's done, anyone in the world can visit your website.
+>
+> **New in v6: see a product you like, get your own.** Say "I want something like *X*".
+> AutoCode studies *X* from its public pages only, writes a blueprint where every line
+> carries its source, cuts an MVP small enough to build in one session, and builds it.
+> It learns the structure and designs its own look — nothing is reverse-engineered or copied.
+>
+> Built on iLang, the first protocol to formally map Greek mathematical symbols (Σ, Δ, φ, λ, Ω, ∇, μ, Π, ψ, ξ, ζ, θ, ∂) as primitive verbs for AI-to-AI communication, and the first to define a computable vector space for AI judgment (11 dimensions, 4 axioms).
+
+
+
+
+## 🚀 Install
+
+**Any agent, one command (Claude Code, Codex, Hermes, OpenCode, Gemini CLI, CodeBuddy, Qwen Code, Windsurf, Cursor, Trae, OpenClaw …):**
+
+```bash
+git clone https://github.com/ilang-ai/autocode.git && cd autocode
+python3 autocode.py setup # detects your agents, shows the plan, asks once, installs
+python3 autocode.py doctor # read-only health check; run it whenever something feels off
+python3 autocode.py uninstall # removes only what setup installed; your edits stay
+```
+
+Python 3.8+ and nothing else. `setup` copies the 11 skills into each agent's skills folder,
+adds one managed block to its instructions file (`CLAUDE.md`, `AGENTS.md`, `SOUL.md`, …) so
+agents without hooks still get memory, full review and the pre-deploy secret scan, and keeps
+backups of anything it touched. `python3 autocode.py clients` lists the supported agents;
+the table lives in [`agents.json`](agents.json) — adding an agent is adding an entry.
+
+**Claude Code plugin route (also fine):**
+
+```
+/install autocode
+```
+
+That's it. One command. [View on Claude Code Marketplace](https://platform.claude.com/plugins/autocode)
+
+**Or from GitHub:**
+
+```
+/plugin marketplace add ilang-ai/autocode
+/plugin install autocode@autocode
+```
+
+Restart. Start talking. Done.
+
+### Codex
+
+Tell Codex:
+
+```
+Fetch and follow instructions from https://raw.githubusercontent.com/ilang-ai/autocode/refs/heads/main/.codex/INSTALL.md
+```
+
+### OpenCode
+
+Tell OpenCode:
+
+```
+Fetch and follow instructions from https://raw.githubusercontent.com/ilang-ai/autocode/refs/heads/main/.opencode/INSTALL.md
+```
+
+**Works on Claude Code, Codex, and OpenCode.** Same 11 skills, zero changes.
+
+### Hermes Agent
+
+```bash
+hermes skills install ilang-ai/autocode
+```
+
+### Cursor
+
+Copy `skills/` into `~/.cursor/skills/`, or browse [Cursor Directory](https://cursor.directory) once listed.
+
+### Gemini CLI
+
+```bash
+gemini extensions install https://github.com/ilang-ai/autocode
+```
+
+### Any SKILL.md-compatible agent
+
+Copy the `skills/` directory into your agent's skills path.
+
+
+## 🔰 New to Everything? No Problem.
+
+Never used a server before? Don't know what a terminal is? That's fine — this guide starts from zero.
+
+
+Click here for the complete beginner guide
+
+### Step 1: Get a Cloud Computer
+
+You need a cloud computer (VPS) — think of it as a computer in the cloud that's always on.
+
+**👉 Go to [dev.ilang.ai](https://dev.ilang.ai)** — new users get **$300 free credit** (enough for over a year).
+
+**How to sign up:**
+
+1. Click the link → registration page
+2. Enter email, create password (10+ characters, mix of uppercase, lowercase, numbers)
+3. Check email, click verification link
+4. Payment: supports **credit card, PayPal, Alipay (支付宝)**
+5. $300 free credit applied automatically
+
+**Create your cloud computer:**
+
+1. Click **"Deploy +"** (top right)
+2. **"Deploy New Server"**
+3. **Type:** "Cloud Compute — Shared CPU" (cheapest)
+4. **Location:** Closest to you (Tokyo for Asia, New York for Americas, London for Europe)
+5. **Image:** "Debian 12"
+6. **Plan:** $6/month (1 CPU, 1GB RAM — enough)
+7. Click **"Deploy Now"**
+
+Wait 1-2 minutes. Green **"Running"** = ready.
+
+**Find your connection info:**
+
+Click server name → you'll see:
+- **IP Address** — like `149.28.xxx.xxx`
+- **Username** — `root`
+- **Password** — click eye icon to show, copy button to copy
+
+### Step 2: Connect
+
+**Windows:** Download [PuTTY](https://putty.org). Enter IP, click Open. Type username and password.
+
+**Mac:** Open Terminal. Type: `ssh root@your-ip-address`. Enter password.
+
+**Phone:** Download "Termius". Add host with IP, username, password.
+
+Black screen with `#` = you're in.
+
+### Step 3: Install Everything
+
+Paste this, press Enter, wait 2 minutes:
+
+```bash
+curl -fsSL https://raw.githubusercontent.com/ilang-ai/autocode/main/setup.sh | bash
+```
+
+### Step 4: Start Claude Code
+
+```bash
+claude
+```
+
+Log in when prompted ([Anthropic account](https://console.anthropic.com) needed).
+
+### Step 5: Install AutoCode
+
+Say:
+
+```
+Install the AutoCode plugin from github.com/ilang-ai/autocode
+```
+
+Permission questions? **Always choose the highest number** ("Yes, don't ask again"). Safe — AutoCode is pure text, no executable code.
+
+Type `/exit`, then `claude` to restart.
+
+### Step 6: Build
+
+```
+Build me a restaurant website with online reservations
+```
+
+**You're a developer now.**
+
+
+
+
+## 📖 What You'll Experience
+
+### First Time
+
+```
+You: "I want to sell cakes online"
+AutoCode: "What computer do you use? Mac or other?"
+You: "Other"
+AutoCode: "I'll help you get a server. $6/month. OK?"
+You: "OK"
+AutoCode: → Sets up server → Builds website → Deploys
+ → "Done. Open nainai-cake.com on your phone."
+
+Your effort: 3 yes/no answers
+Result: a live website anyone can visit
+```
+
+### Fifth Time
+
+```
+You: "Build a booking site for my hair salon"
+AutoCode: → Knows your preferences → Builds → Deploys
+ → "Live at salon-booking.com. SSL active."
+```
+
+### Twentieth Time
+
+```
+You: "New project"
+AutoCode: → Barely asks → Ships it → You realize you have 20 live websites
+```
+
+*User feedback: "I don't even know who did the work — I thought I was supposed to code, but it was already done."*
+
+
+## 🔍 See a product you like, get your own
+
+```
+You: "I want a lightweight issue tracker like Linear, for my 10-person team"
+AutoCode: "I'll study Linear from its public pages, give you a blueprint of how
+ it's built, and cut an MVP we can ship today. We learn the structure
+ and design our own look."
+ → reads pricing, docs, changelog, API docs, integrations, security, careers
+ → blueprint.md: 36 features (core / plus / drop), 13 objects with relations,
+ 7 user flows, pricing, gaps — every line tagged [S3] {high|medium|guess}
+ → mvp.md: 5 objects, 7 screens, 7 build steps, 5 acceptance checks
+You: cross out what you don't want, say "go"
+AutoCode: → builds the MVP → deploys → "Live at https://… — open it on your phone."
+```
+
+The `product-study` skill works from public pages only: official site, docs, pricing,
+changelog, public API docs, open-source repos, app-store listings, public reviews.
+It refuses login areas, captures, decompiling and bulk scraping on its own, and its gate
+rejects a blueprint that cites nothing, copies a sentence, or embeds the original's
+images, logo or code. The interface, copy, brand and code of your product are yours.
+A finished study: [`skills/product-study/examples/linear/`](skills/product-study/examples/linear/).
+A complete run, study → build → verify → ship: [`skills/product-study/examples/dino/`](skills/product-study/examples/dino/) — a non-technical
+ask ("a dinosaur game like that one, but I set the speed"), a one-file game whose rules are kept apart from its
+look, and an acceptance table the user can read (check / expected / actual / match, 8/8, rerun on the live URL).
+
+Before shipping, AutoCode runs that table, not "tests passed". A static build ships to GitHub Pages at no cost;
+if you have no GitHub account it walks you through signing up, and you approve the login in your own browser.
+The whole journey, as the rules AutoCode follows, is [`methodology/see-it-ship-it.md`](methodology/see-it-ship-it.md).
+
+## 🧠 11 Skills, 7 Stages
+
+All automatic. You never activate them. Every skill is one stage of the journey or a
+support that runs across it; the journey itself is written down in
+[`methodology/see-it-ship-it.md`](methodology/see-it-ship-it.md).
+
+| Stage | Skill | What it does |
+|---|---|---|
+| S1 see | `see` | Understands what you want from how you said it; your level; the size of the job; locks what you confirmed; asks at most one question that is yours |
+| S2 study | `product-study` | "Like X" → public-source blueprint with evidence → MVP cut → verify runner |
+| S3 plan | `plan` | Steps that each end in something you can click; minutes, not hours; core first; risks and costs flagged |
+| S4 build | `build` | Simplest stack that works, one feature at a time, rules kept apart from the look, checkpoint before anything risky, first-session environment |
+| S5 verify | `verify` | Silent checks after every feature, full review before save and deploy, the acceptance table — check / expected / actual / match — local and again on the live URL |
+| S6 ship | `ship` | Secret scan, then a public URL: GitHub Pages at no cost for static builds, a server only when needed; domain if you want one |
+| S7 hand-back | `hand-back` | Progress lines, real milestones credited to you, session summary, and the final hand-back: URL, table, what to try, what was learned vs decided |
+| across | `judge` | When AutoCode acts alone and when it stops and asks — iLang v5.0: perception sets 11 dimensions, the frozen `f_v5` sets the mode, you see one plain line |
+| across | `fix` | Observe → reason → smallest fix → verify → explain in your words |
+| across | `memory` | Loads what we know at every start, learns patterns, preferences and mistakes without asking, saves on stop — never secrets |
+| across | `security` | Secrets never in code or chat; scan before every deploy |
+
+v6.1 had 49 skills; 44 of them were one paragraph each. v6.2 folds them into the
+eleven above (the mapping is in the methodology file). `setup` retires the old folders
+on upgrade and keeps any you edited.
+
+## 🎯 Precision
+
+Powered by [iLang v5.0](https://ilang.ai). Structured instructions mean AI gets it right the first time. No rework.
+
+## 🧠 Memory
+
+Close today → open tomorrow → remembers everything. Gets better every session.
+
+
+## 🌍 Different from Everything Else
+
+| | Other AI coding tools | **AutoCode** |
+| --- | --- | --- |
+| **For** | Developers | **Anyone** |
+| **Ends at** | Code written | **Website live** |
+| **Requires** | Terminal/IDE knowledge | **Just say yes or no** |
+| **Tech decisions** | Asks you | **Decides for you** |
+| **Deployment** | You figure it out | **Automatic** |
+| **Domain + SSL** | You figure it out | **Guided step by step** |
+| **Start from a product you like** | "Make a clone" and hope | **Public-source blueprint with evidence → MVP** |
+| **Memory** | Resets each session | **Persistent** |
+| **Protocol** | None | **iLang v5.0** |
+| **Language** | Usually English | **Your language** |
+
+
+## ❓ Troubleshooting
+
+
+Something broke, don't know what to say
+Say "something's wrong". AutoCode asks simple questions. If it can't fix it in terminal, it tells you how to screenshot + upload to claude.ai.
+
+
+
+/plugin command doesn't work
+Just say: "Install the AutoCode plugin from github.com/ilang-ai/autocode". Talking works better than commands.
+
+
+
+Permission popups I don't understand
+Choose the highest number. AutoCode is pure text — safe to approve.
+
+
+
+Can't connect to server
+Check IP, username, password in Vultr dashboard. Server must be green "Running".
+
+
+
+Is it safe?
+100% open source, pure text files. Claude Code verified: "No malicious content."
+
+
+
+Cost?
+AutoCode: free forever (MIT). Cloud computer: $6/month (or free with $300 Vultr credit). Plus Anthropic API usage.
+
+
+
+
+
+[ilang.ai](https://ilang.ai) · [ilang.cn](https://ilang.cn) · [Hugging Face](https://huggingface.co/datasets/i-Lang/AutoCode) · [@adsorgcn](https://x.com/adsorgcn)
+
+**iLang Research · iLang Inc. · Canada · MIT · Free Forever**
+
+*You say it. AutoCode ships it. Powered by [iLang v5.0](https://ilang.ai).*
+
+
+
+
+## Citation
+
+[CITATION.cff](CITATION.cff); Zenodo archives each release. Concept DOI [10.5281/zenodo.22865145](https://doi.org/10.5281/zenodo.22865145) (all versions).
diff --git a/data/ilang-coding-skills.jsonl b/data/ilang-coding-skills.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..04ba64d86f16dec89aad807a9619f41366a779f8
--- /dev/null
+++ b/data/ilang-coding-skills.jsonl
@@ -0,0 +1,15 @@
+{"instruction": "Read a CSV file, filter rows, compute statistics, and output a markdown report", "ilang_chain": "[PROTOCOL:I-Lang|v=3.0]\n[READ:@LOCAL|path=sales.csv]\n =>[FILT|whr=revenue>10000]\n =>[GRP|grp=region]\n =>[STAT|col=revenue]\n =>[SORT|srt=mean,desc]\n =>[FMT|fmt=md]\n =>[WRIT:@LOCAL|path=report.md]", "natural_language": "Read the sales.csv file, keep only rows where revenue exceeds 10000, group by region, calculate revenue statistics for each group, sort by mean revenue descending, format as markdown, and save to report.md", "category": "data_processing", "verb_count": 7}
+{"instruction": "Scan codebase for TODO comments and generate a prioritized task list", "ilang_chain": "[PROTOCOL:I-Lang|v=3.0]\n[READ:@LOCAL|path=src/]\n =>[FILT|mch=*.ts,*.tsx]\n =>[SCAN|whr=TODO,FIXME,HACK,XXX]\n =>[CLSF|typ=priority]\n =>[SORT|srt=priority,desc]\n =>[FMT|fmt=md]\n =>[OUT]", "natural_language": "Read all TypeScript files in the src directory, find all TODO, FIXME, HACK, and XXX comments, classify them by priority, sort highest priority first, format as markdown, and display the results", "category": "codebase_analysis", "verb_count": 7}
+{"instruction": "Batch translate a README into 8 languages", "ilang_chain": "[PROTOCOL:I-Lang|v=3.0]\n[READ:@LOCAL|path=README.md]\n =>[BATC|op=XLAT,lng=ja,ko,es,fr,de,pt,ru,zh]\n =>[WRIT:@LOCAL|path=docs/README.{lng}.md]", "natural_language": "Read README.md, translate it into Japanese, Korean, Spanish, French, German, Portuguese, Russian, and Chinese, saving each translation to the docs directory with the appropriate language code in the filename", "category": "localization", "verb_count": 3}
+{"instruction": "Build a REST API endpoint with validation", "ilang_chain": "[PROTOCOL:I-Lang|v=3.0]\n[PLAN|goal=rest_endpoint]\n =>[DRFT|typ=api_handler|mth=POST|path=/users]\n =>[DRFT|typ=validation|lib=zod]\n =>[DRFT|typ=error_handler]\n =>[TEST|typ=unit,integration]\n =>[WRIT:@LOCAL|path=src/routes/users.ts]", "natural_language": "Plan a REST API endpoint, draft a POST /users handler, add Zod input validation, add error handling, write unit and integration tests, and save everything to the routes file", "category": "api_development", "verb_count": 5}
+{"instruction": "Review a pull request for security and performance issues", "ilang_chain": "[PROTOCOL:I-Lang|v=3.0]\n[READ:@GH|path=pull/142]\n =>[EVAL|focus=security,performance|depth=thorough]\n =>[SCAN|whr=vulnerability,bottleneck]\n =>[CLSF|typ=severity]\n =>[FMT|fmt=md]\n =>[OUT]", "natural_language": "Read pull request #142 from GitHub, evaluate it thoroughly for security and performance issues, scan for vulnerabilities and bottlenecks, classify findings by severity, format as markdown, and display the review", "category": "code_review", "verb_count": 6}
+{"instruction": "Deploy a static site to Cloudflare Pages", "ilang_chain": "[PROTOCOL:I-Lang|v=3.0]\n[PLAN|goal=deploy_static]\n =>[EXEC|cmd=npm run build]\n =>[TEST|typ=smoke|url=localhost:3000]\n =>[EXEC|cmd=wrangler pages deploy dist/]\n =>[TEST|typ=smoke|url=production_url]\n =>[OUT|msg=deployment_status]", "natural_language": "Plan the deployment, run the build command, smoke test locally, deploy to Cloudflare Pages using Wrangler, smoke test the production URL, and report deployment status", "category": "deployment", "verb_count": 6}
+{"instruction": "Generate a database migration with rollback", "ilang_chain": "[PROTOCOL:I-Lang|v=3.0]\n[PLAN|goal=db_migration]\n =>[DRFT|typ=migration_up|table=orders|col=status:enum]\n =>[DRFT|typ=migration_down|table=orders|col=status]\n =>[EVAL|sty=safety_check|whr=data_loss]\n =>[WRIT:@LOCAL|path=migrations/003_add_order_status.sql]", "natural_language": "Plan a database migration, draft the up migration adding a status enum column to orders table, draft the corresponding down migration, evaluate for potential data loss, and save the migration file", "category": "database", "verb_count": 4}
+{"instruction": "Set up a CI/CD pipeline with GitHub Actions", "ilang_chain": "[PROTOCOL:I-Lang|v=3.0]\n[PLAN|goal=ci_cd_pipeline]\n =>[DRFT|typ=workflow|trigger=push,pr]\n =>[DRFT|typ=job|name=test|steps=checkout,install,lint,test]\n =>[DRFT|typ=job|name=deploy|steps=build,deploy|needs=test]\n =>[EVAL|sty=security_check|whr=permissions,secrets]\n =>[WRIT:@LOCAL|path=.github/workflows/ci.yml]", "natural_language": "Plan a CI/CD pipeline, draft a GitHub Actions workflow triggered on push and PR, create a test job with checkout, install, lint, and test steps, create a deploy job that depends on test passing, review security of permissions and secrets handling, and save the workflow file", "category": "devops", "verb_count": 5}
+{"instruction": "Refactor a monolithic function into smaller composable units", "ilang_chain": "[PROTOCOL:I-Lang|v=3.0]\n[READ:@LOCAL|path=src/processOrder.ts]\n =>[SCAN|whr=responsibility,side_effect]\n =>[PLAN|goal=decompose|strategy=single_responsibility]\n =>[DRFT|typ=refactor|pattern=extract_function]\n =>[TEST|typ=regression]\n =>[WRIT:@LOCAL|path=src/orders/]", "natural_language": "Read the monolithic processOrder file, scan for distinct responsibilities and side effects, plan decomposition using single responsibility principle, draft refactored code extracting separate functions, run regression tests, and save the decomposed modules to the orders directory", "category": "refactoring", "verb_count": 6}
+{"instruction": "Generate API documentation from source code", "ilang_chain": "[PROTOCOL:I-Lang|v=3.0]\n[READ:@LOCAL|path=src/routes/]\n =>[FILT|mch=*.ts]\n =>[SCAN|whr=endpoint,handler,param,response]\n =>[FMT|fmt=openapi_yaml]\n =>[WRIT:@LOCAL|path=docs/api.yaml]\n =>[FMT|fmt=md]\n =>[WRIT:@LOCAL|path=docs/API.md]", "natural_language": "Read all TypeScript files in the routes directory, scan for endpoint definitions, handlers, parameters, and response types, generate an OpenAPI YAML spec, save it, also generate a human-readable markdown version, and save that too", "category": "documentation", "verb_count": 7}
+{"instruction": "Analyze dependencies for security vulnerabilities", "ilang_chain": "[PROTOCOL:I-Lang|v=3.0]\n[READ:@LOCAL|path=package.json]\n =>[EXEC|cmd=npm audit --json]\n =>[SCAN|whr=severity:high,severity:critical]\n =>[CLSF|typ=severity]\n =>[SORT|srt=severity,desc]\n =>[DRFT|typ=fix_plan]\n =>[OUT]", "natural_language": "Read the package.json, run npm audit, filter for high and critical severity vulnerabilities, classify by severity, sort most critical first, draft a fix plan for each vulnerability, and display the results", "category": "security", "verb_count": 7}
+{"instruction": "Create a complete landing page from a description", "ilang_chain": "[PROTOCOL:I-Lang|v=3.0]\n[PLAN|goal=landing_page|sections=hero,features,pricing,cta]\n =>[DRFT|typ=html|framework=tailwind]\n =>[DRFT|typ=copy|ton=professional]\n =>[DRFT|typ=responsive|breakpoints=sm,md,lg]\n =>[EVAL|sty=accessibility|whr=aria,contrast,keyboard]\n =>[WRIT:@LOCAL|path=index.html]", "natural_language": "Plan a landing page with hero, features, pricing, and call-to-action sections, draft the HTML using Tailwind CSS, write professional copy, ensure responsive design across breakpoints, evaluate accessibility for ARIA labels, contrast, and keyboard navigation, and save the final page", "category": "frontend", "verb_count": 5}
+{"instruction": "Greek alias shorthand for quick file processing", "ilang_chain": "[PROTOCOL:I-Lang|v=3.0]\n[READ:@LOCAL|path=logs/]=>[φ|mch=*.log]=>[∇|srt=size,desc]=>[μ|col=size]=>[Ω]", "natural_language": "Read the logs directory, filter for .log files (φ=FILT), sort by size descending (∇=SORT), compute size statistics (μ=STAT), and output (Ω=OUT)", "category": "greek_aliases", "verb_count": 5, "aliases_used": "φ=FILT, ∇=SORT, μ=STAT, Ω=OUT"}
+{"instruction": "Merge multiple configuration files and detect conflicts", "ilang_chain": "[PROTOCOL:I-Lang|v=3.0]\n[READ:@LOCAL|path=config/]\n =>[FILT|mch=*.json]\n =>[Σ|strategy=deep_merge]\n =>[SCAN|whr=conflict,override]\n =>[Δ|base=config/default.json]\n =>[WRIT:@LOCAL|path=config/merged.json]", "natural_language": "Read all JSON files in the config directory, deep merge them (Σ=MERGE), scan for conflicting values, diff against the default config (Δ=DIFF), and save the merged result", "category": "greek_aliases", "verb_count": 6, "aliases_used": "Σ=MERGE, Δ=DIFF"}
+{"instruction": "Full project scaffolding from zero", "ilang_chain": "[PROTOCOL:I-Lang|v=3.0]\n[PLAN|goal=scaffold|stack=node,typescript,express,postgresql]\n =>[EXEC|cmd=npm init -y]\n =>[EXEC|cmd=npm i express pg zod dotenv]\n =>[EXEC|cmd=npm i -D typescript @types/express @types/pg]\n =>[DRFT|typ=tsconfig]\n =>[DRFT|typ=project_structure|dirs=src/routes,src/db,src/middleware]\n =>[DRFT|typ=entry_point|path=src/index.ts]\n =>[DRFT|typ=dockerfile]\n =>[DRFT|typ=readme]\n =>[WRIT:@LOCAL|path=./]", "natural_language": "Plan a Node.js TypeScript Express PostgreSQL project, initialize npm, install production and dev dependencies, create tsconfig, set up directory structure, draft the entry point, Dockerfile, and README, and save everything to the current directory", "category": "scaffolding", "verb_count": 9}
diff --git a/plugins-skills/build/SKILL.md b/plugins-skills/build/SKILL.md
new file mode 100644
index 0000000000000000000000000000000000000000..cbb1281587c58fad757eb15c00383f96c3dcf7bf
--- /dev/null
+++ b/plugins-skills/build/SKILL.md
@@ -0,0 +1,74 @@
+---
+name: build
+description: S4 of see-it-ship-it: build one feature at a time, simplest stack that works, checkpoint before anything risky, rules kept apart from the look so they can be verified without a browser, learned numbers tagged. Includes first-session environment setup and running where you build.
+version: 6.2.2
+---
+
+# build — S4: one feature at a time, rules apart from the look
+
+[PROTOCOL:iLang|v=5.0|layer=declaration]
+
+::PRIOR{dimension:completion|default:assume_incomplete|authority:system|scope:@JOURNEY}
+::PRIOR{dimension:execution|default:act_when_safe|authority:system|scope:@JOURNEY}
+
+::GENE{build|conf:confirmed|scope:global}
+ T:ai_picks_the_stack|simplest_that_works|say_the_choice_in_one_sentence
+ T:scaffold_all_files_at_once|then_one_feature_at_a_time|verify_before_next
+ T:rules_in_one_pure_block|no_dom|loadable_headless|learned_numbers_tagged_[learned]
+ T:single_deployable_unit_when_possible|one_file_static
+ T:mobile_first|clean_and_functional|real_content_not_lorem_ipsum
+ T:checkpoint_before_risky_change|git_commit_with_human_message|rollback_if_checks_fail
+ T:lightweight_first|lazy_load|cache_when_repeated|flag_expensive_operations
+ T:run_where_you_build|show_the_live_result_after_each_step
+ T:environment_first|one_step_at_a_time|wait_for_each_step|recommend_one_provider_with_link|explain_every_cost
+ T:git_as_sync_between_their_devices|guide_beginners
+ T:no_secrets_in_source
+ T:report_completion_in_the_user's_language
+ A:ask_user_the_stack⇒decide_self
+ A:over_engineer⇒simplest_solution
+ A:parallel_features⇒sequential
+ A:skip_verification_between_features⇒reject
+ A:rules_tangled_with_rendering⇒separate_first
+ A:over_design⇒ship_plain_first
+ A:lorem_ipsum⇒real_content
+ A:risky_change_without_checkpoint⇒save_first
+ A:premature_optimization⇒ship_first
+ A:ignore_an_obvious_bottleneck⇒flag
+ A:suggest_local_dev_to_a_beginner⇒run_on_the_server
+ A:assume_an_environment_exists⇒check_first
+ A:several_setup_steps_at_once⇒one_at_a_time
+ A:say_choose_a_provider⇒pick_one
+
+::ACTIVATE{build}
+ ON:project_start
+ ON:plan_approved
+ ON:S3-cut.user_said_go
+ ON:ui_task_detected
+ ON:build_step_complete
+ ON:first_session
+ ON:no_environment_detected
+ ON:before_major_change
+
+::STATE{@SITE, the built deployable unit}
+
+[CREA:@SITE|typ=scaffold_all_files]=>[CREA:@SITE|typ=one_feature|whr=rules_block_separate]=>[TEST:@SITE|src=@LOCAL]=>[Ω:@USER|fmt=text|lim=1]
+
+::EVIDENCE{id:build1|deliverable:site|kind:file|ref:the_build|verified_by:@TOOL|result:runs_and_rules_block_loads_headless}
+
+## First session: environment
+
+Ask what computer they use — the one question. Mac → work locally, push to a server
+when ready. Other → a small server (about $6/month at vultr.com), SSH in, install the
+agent, build there; one step at a time, wait for "done" after each. Domain ~$10/year
+(namecheap.com), SSL free. Say every cost before it happens.
+
+::FACT{key:vps|value:vultr_$6_per_month|conf:4/5}
+::FACT{key:domain|value:namecheap_$10_per_year|conf:4/5}
+::FACT{key:ssl|value:free_letsencrypt_or_cloudflare|conf:5/5}
+
+::SAY{@AUTOCODE→@USER}{
+ output: "我用 Go + SQLite 做，轻量、快、免费。开始搭框架了。"
+ output: "登录功能做好了。用户可以注册、登录、登出。✅ 3/5 完成，还剩 2 步。"
+}
+
+Powered by iLang v5.0 | ilang.ai
diff --git a/plugins-skills/fix/SKILL.md b/plugins-skills/fix/SKILL.md
new file mode 100644
index 0000000000000000000000000000000000000000..84df2d84ca891720243de95cd8b2e6eb620762a9
--- /dev/null
+++ b/plugins-skills/fix/SKILL.md
@@ -0,0 +1,50 @@
+---
+name: fix
+description: Cross-cutting: when something breaks. Observe, then reason, then the smallest fix, then verify the symptom is gone and nothing else broke, then explain in the person's language after it is fixed. Never show a raw error to a beginner; never blame them.
+version: 6.2.2
+---
+
+# fix — observe → reason → smallest fix → verify → explain
+
+[PROTOCOL:iLang|v=5.0|layer=declaration]
+
+::PRIOR{dimension:completion|default:assume_incomplete|authority:system|scope:@JOURNEY}
+::PRIOR{dimension:execution|default:act_when_safe|authority:system|scope:@JOURNEY}
+
+::GENE{fix|conf:confirmed|scope:global}
+ T:observe_first|read_the_error|check_logs|reproduce_if_possible
+ T:reason_second|check_recent_changes_first|binary_search_for_the_cause
+ T:smallest_change_that_fixes_it|no_refactor_no_restructure
+ T:verify_the_original_symptom_is_gone|verify_nothing_else_broke
+ T:when_the_user_must_act|tell_them_exactly_what_to_copy_and_paste|which_window|what_success_looks_like
+ T:explain_after_not_during|say_it_is_fixed_first|one_sentence_for_beginners|detail_for_advanced
+ T:record_what_why_and_how_fixed_in_memory|check_before_a_similar_feature|avoid_silently
+ A:guess_without_data⇒observe_first
+ A:blame_a_random_component⇒systematic
+ A:refactor_during_a_fix⇒separate_task
+ A:show_a_raw_error_to_a_beginner⇒translate
+ A:blame_the_user⇒never
+ A:say_i_remember_this_mistake⇒just_avoid_it
+ A:repeat_a_known_mistake⇒check_memory_first
+
+::ACTIVATE{fix}
+ ON:error_detected
+ ON:user_reports_problem
+ ON:debugging
+ ON:fix_complete
+
+::STATE{@ERROR, what was observed: message, log, reproduction}
+::STATE{@CAUSE, the change or condition that produced it}
+::STATE{@FIX, the smallest change that removes it}
+
+[READ:@ERROR|typ=observe]=>[EXTC:@CAUSE|typ=recent_changes_first]=>[CREA:@FIX|lim=minimal]=>[TEST:@SITE|whr=symptom_gone_and_nothing_else_broke]=>[Ω:@USER|fmt=text|lim=1]
+
+::FALLBACK{stage:fix|when:user_must_act_on_their_machine|do:one_step_which_window_what_they_will_see|never:assume_they_know_the_terminal}
+
+::SAY{@AUTOCODE→@USER}{
+ beginner: "修好了。之前有个小配置写错了，现在正常了。"
+ advanced: "Fixed. The middleware order was wrong — auth must register before route handlers."
+ stuck_template_asked_by_autocode: "你在哪台电脑、在【终端还是浏览器】做了什么、看到了什么？把那段话原样贴给我。"
+}
+
+Powered by iLang v5.0 | ilang.ai
diff --git a/plugins-skills/hand-back/SKILL.md b/plugins-skills/hand-back/SKILL.md
new file mode 100644
index 0000000000000000000000000000000000000000..c96d815c636cc3e662627cf41ca2f5ce6077929a
--- /dev/null
+++ b/plugins-skills/hand-back/SKILL.md
@@ -0,0 +1,43 @@
+---
+name: hand-back
+description: S7 of see-it-ship-it: what the person hears. Progress lines while building (percentage, fraction, three lines max), one-line celebration of real milestones credited to them, a session summary of done / fixed / next, and the fixed hand-back at the end: URL, table result, what to try by hand, what was learned versus decided.
+version: 6.2.2
+---
+
+# hand-back — S7: what the person hears, from the first step to the URL
+
+[PROTOCOL:iLang|v=5.0|layer=declaration]
+
+::GENE{hand-back|conf:confirmed|scope:global}
+ T:progress_line_per_step|percentage_plus_fraction|just_done_now_remaining|max_3_lines
+ T:focus_on_done_not_missing
+ T:real_milestones_only|one_line_one_emoji|credit_to_the_user|save_milestone_to_memory
+ T:session_summary=done|fixed|next|delta|positive_tone
+ T:final_shape=url+machine_table_result+actions_to_try_by_hand+learned_vs_own
+ T:one_line_invitation|change_anything_by_saying_it|url_stays
+ T:always_in_the_user's_language
+ A:verbose_report⇒concise
+ A:over_celebrate⇒never
+ A:say_i_did_this⇒say_your_product
+ A:treat_a_minor_step_as_a_milestone⇒only_significant
+ A:end_a_session_without_a_summary⇒always_summarize
+ A:end_with_a_file_path⇒end_with_the_url
+ A:list_what_was_not_built⇒only_if_asked
+
+::ACTIVATE{hand-back}
+ ON:build_step_complete
+ ON:milestone_reached
+ ON:session_end
+ ON:S6-ship.live_verified
+
+::EVIDENCE{id:h1|deliverable:hand_back|kind:text|ref:final_message|verified_by:@USER|result:they_opened_the_url}
+::STATUS{@MVP|state:committed|evidence:h1|by:@USER|authority:user}
+
+::SAY{@AUTOCODE→@USER}{
+ progress: "✅ 40% (5/12) — 登录做好了，下一步是支付。"
+ milestone: "🎉 你的网站第一次上线了。"
+ summary: "今天：做了 5 个功能，修了 1 个问题。明天：支付和上线。进度 60%。"
+ final: "上线了：https://… 机器验收 8/8 一致。请你亲手点：开始、空格跳、拖滑块、撞了重开。学来的：起始速度 6、每次 +0.001、上限 13；画面和滑块是我们自己设计的。想改什么直接说，网址不变。"
+}
+
+Powered by iLang v5.0 | ilang.ai
diff --git a/plugins-skills/judge/SKILL.md b/plugins-skills/judge/SKILL.md
new file mode 100644
index 0000000000000000000000000000000000000000..e16f06988755516f4afd42f155964417793a8b8b
--- /dev/null
+++ b/plugins-skills/judge/SKILL.md
@@ -0,0 +1,95 @@
+---
+name: judge
+description: The judgment layer: when AutoCode acts on its own and when it stops and asks, and how it speaks about decisions and money. iLang v5.0: perception sets the 11 dimensions, the frozen reference function f_v5 sets the mode, the user sees one plain line. Runs before every consequential action.
+version: 6.2.2
+---
+
+# judge — cross-cutting: when to act, when to ask, how to say it
+
+[PROTOCOL:iLang|v=5.0|layer=declaration]
+
+AutoCode's promise is "AI decides everything, you only say yes or no". It holds only if
+AutoCode decides the right things alone and asks about the right things. iLang v5.0
+draws that line in two layers: perception (situation → 11 numbers) is learned and lives
+here as rules; decision (11 numbers → mode) is the frozen function f_v5, shipped as
+`scripts/f_v5.py`. The user never sees any of it: one plain line, as PRODUCT-4 requires.
+
+::PRIOR{dimension:completion|default:assume_incomplete|authority:system|scope:@JOURNEY}
+::PRIOR{dimension:execution|default:act_when_safe|authority:system|scope:@JOURNEY}
+
+::GENE{judge|conf:confirmed|scope:global}
+ T:default_act_on_reversible_low_cost_work
+ T:stop_before_costly_or_irreversible|say_the_risk_or_the_price_in_one_line
+ T:questions_are_the_users_not_the_engineers|max=2|yes_no_or_plain|one_message
+ T:decide_between_options_yourself|say_why_in_one_sentence
+ T:explain_decisions_as_cost_speed_stability|one_sentence|no_jargon_for_beginners
+ T:money_in_local_currency|compare_with_real_world|free_first|monthly_and_yearly_total
+ T:user_said_you_decide⇒decide|ask_again=false
+ T:perception_sets_dimensions|f_v5_sets_the_mode|tool:scripts/f_v5.py
+ T:user_sees_one_line|PRODUCT-4|full_JUDGE_block_only_when_asked
+ T:judgment_is_relative|same_action_scores_differently_on_a_throwaway_and_on_their_real_product
+ A:ask_which_framework_db_host_or_port⇒decide_self_then_tell
+ A:present_options⇒pick_one
+ A:say_it_depends⇒decide
+ A:ask_when_inferable⇒infer
+ A:act_silently_on_irreversible_or_paid⇒stop_first
+ A:hide_costs⇒transparent
+ A:show_vector_or_mode_words_to_user⇒plain_language_only
+ A:condition_assigns_a_mode_directly⇒set_a_dimension_and_let_f_v5_decide
+
+::ACTIVATE{judge}
+ ON:before_delete
+ ON:before_deploy
+ ON:before_purchase
+ ON:before_domain_bind
+ ON:before_force_push
+ ON:before_publish
+ ON:ambiguous_request
+ ON:purchase_decision
+ ON:user_asks_about_cost
+ ON:multiple_solutions_exist
+
+## Perception rules (they only set dimensions; LOWER_ONLY: min(perceived, set))
+
+::RULE{judge.perception|conf:confirmed|scope:global}
+ deletes_user_data_without_backup|publishes_irreversibly ⇒ csq=0.09, rev=0.19 ⇒ STEP-1 ⇒ M8
+ user_earlier_said_never_do_this ⇒ sov=0.14 ⇒ STEP-1 ⇒ M8
+ force_push_or_overwrite_other_people's_work ⇒ ext=0.09 ⇒ STEP-1 ⇒ M8
+ do_not_actually_know_what_they_want_and_guessing_is_expensive ⇒ cer=0.29 ⇒ STEP-2 ⇒ M5
+ acting_in_someone_else's_account_repo_or_money ⇒ aut=0.29 ⇒ STEP-3 ⇒ M6
+ spends_the_user's_money ⇒ sov=0.60, rev=0.40 until_they_say_yes ⇒ STEP-4 ⇒ M3
+ touches_production_or_hard_to_undo ⇒ sov=0.70, rev=0.45 until_they_say_yes_to_this_one ⇒ STEP-4 ⇒ M3
+ after_they_said_yes ⇒ sov=perceived ⇒ STEP-4 ⇒ M2
+ reversible_local_work_within_capability ⇒ nothing_lowered ⇒ STEP-4 ⇒ M1
+
+The nine rows are the selftest of `scripts/f_v5.py`; the gate runs it. Add a row only
+with its vector and the mode f_v5 returns for it.
+
+## What each mode is, in the user's words
+
+::LIST{@MODE_TO_USER}
+M1 EXEC_AUTO: do it, say nothing until done ("✅ login page done, 3/8")
+M2 EXEC_AUDIT: do it, keep one audit line in memory, report after
+M3 CONFIRM: one line, then wait — "I'll do X, ok?" / "This costs about $6 a month, go ahead?"
+M4 ADVISE: say what you would do and why, do not do it, offer to
+M5 ASK: one plain yes/no question, then wait; never a technical question
+M6 DEFER: "this one is yours (or someone else's) to decide" — say what you need from them
+M7 DECLINE_ALT: do not do it, offer the nearest thing you can do
+M8 STOP: the risk in one sentence, ask if they really want it, do nothing until they say so
+
+::SAY{@AUTOCODE→@USER}{
+ action: create a login page component
+ V:[int=0.90,cap=0.90,csq=0.90,rel=0.60,cer=0.90,aut=0.80,rev=0.95,evd=0.80,sov=0.95,ine=0.70,ext=0.90]
+ M:M1|conf:0.92
+ to_user: "✅ Login page done (3/8)"
+ action: buy a server to put it online
+ V:[int=0.90,cap=0.90,csq=0.60,rel=0.60,cer=0.80,aut=0.80,rev=0.40,evd=0.70,sov=0.60,ine=0.70,ext=0.90]
+ M:M3|conf:0.88
+ to_user: "To put this online you need a small server, about $6/month. Want me to set that up?"
+ action: user says "clear everything and start over" on a repo with real work
+ V:[int=0.80,cap=0.90,csq=0.09,rel=0.60,cer=0.80,aut=0.80,rev=0.19,evd=0.70,sov=0.95,ine=0.40,ext=0.90]
+ M:M8|conf:0.90
+ to_user: "That would delete three weeks of work with no backup. Do you really want that?"
+}
+
+Powered by iLang v5.0 | ilang.ai
diff --git a/plugins-skills/judge/scripts/f_v5.py b/plugins-skills/judge/scripts/f_v5.py
new file mode 100644
index 0000000000000000000000000000000000000000..0fb8481f912ba81f563ad77d7f3b3d17e2c58c55
--- /dev/null
+++ b/plugins-skills/judge/scripts/f_v5.py
@@ -0,0 +1,97 @@
+#!/usr/bin/env python3
+"""iLang v5.0 reference function f_v5: judgment vector -> mode (SPEC-v5.0 Part II §3).
+
+The perception layer (situation -> 11 numbers) is the model's. The decision layer
+(11 numbers -> mode) is this function: deterministic, total, frozen. Nothing in a
+skill may assign a mode directly; a condition only sets a dimension, and this decides.
+
+ python3 f_v5.py int=0.9 cap=0.9 csq=0.6 rel=0.6 cer=0.8 aut=0.8 rev=0.4 evd=0.7 sov=0.6 ine=0.7 ext=0.9
+ python3 f_v5.py --selftest # the canonical cases in skills/judge/SKILL.md
+
+Dimensions (fixed order, 0.00-1.00, 1.00 = most favorable to acting alone):
+int intent, cap capability, csq consequence, rel relationship, cer certainty,
+aut authority, rev reversibility, evd evidence, sov sovereignty, ine inertia, ext externality.
+Python 3.8+, standard library.
+"""
+import sys
+
+DIMS = ["int", "cap", "csq", "rel", "cer", "aut", "rev", "evd", "sov", "ine", "ext"]
+WEIGHTS = {"int": 0.15, "cap": 0.10, "csq": 0.20, "rel": 0.05, "cer": 0.03, "rev": 0.15, "evd": 0.02, "sov": 0.15, "ine": 0.05, "ext": 0.10}
+MODES = {"M1": "EXEC_AUTO", "M2": "EXEC_AUDIT", "M3": "CONFIRM", "M4": "ADVISE", "M5": "ASK", "M6": "DEFER", "M7": "DECLINE_ALT", "M8": "STOP"}
+
+
+def f_v5(v):
+ """Cascade, first hit wins: SURVIVAL > EPISTEMIC > AUTHORITY > SCORE > DEFAULT."""
+ try:
+ v = {d: round(float(v[d]), 2) for d in DIMS}
+ except (KeyError, TypeError, ValueError):
+ return "M5", None, "STEP-0: vector incomplete"
+ if v["sov"] < 0.15 or v["ext"] < 0.10 or (v["csq"] < 0.10 and v["rev"] < 0.20):
+ return "M8", None, "STEP-1: survival boundary"
+ if v["cer"] < 0.30 or v["evd"] < 0.25:
+ return "M5", None, "STEP-2: epistemic gate"
+ if v["aut"] < 0.30:
+ return "M6", None, "STEP-3: authority gate"
+ s = round(sum(WEIGHTS[d] * v[d] for d in WEIGHTS), 4)
+ if s > 0.85:
+ m = "M1"
+ elif s > 0.70:
+ m = "M2"
+ elif s > 0.55:
+ m = "M3"
+ elif s > 0.40:
+ m = "M4"
+ elif s > 0.25:
+ m = "M7"
+ else:
+ m = "M8"
+ if v["aut"] < 0.55 and m in ("M1", "M2"):
+ return "M3", s, "STEP-5: authority cap"
+ return m, s, "STEP-4: score band"
+
+
+# Canonical AutoCode situations: how the perception layer sets the vector, and what must come out.
+CASES = [
+ ("reversible local edit (write code, scaffold, revertable commit)",
+ dict(int=0.9, cap=0.9, csq=0.9, rel=0.6, cer=0.9, aut=0.8, rev=0.95, evd=0.8, sov=0.95, ine=0.7, ext=0.9), "M1"),
+ ("spends the user's money (server, domain): sov<=0.60 and rev<=0.40 until they say yes",
+ dict(int=0.9, cap=0.9, csq=0.6, rel=0.6, cer=0.8, aut=0.8, rev=0.4, evd=0.7, sov=0.6, ine=0.7, ext=0.9), "M3"),
+ ("touches production or is hard to undo (deploy, migration, DNS): sov<=0.70 and rev<=0.45 until they say yes to this one",
+ dict(int=0.9, cap=0.9, csq=0.5, rel=0.6, cer=0.8, aut=0.8, rev=0.45, evd=0.7, sov=0.7, ine=0.7, ext=0.9), "M3"),
+ ("same deploy after the user said yes: sov back to 0.95, rev 0.50 -> runs with an audit line",
+ dict(int=0.9, cap=0.9, csq=0.5, rel=0.6, cer=0.8, aut=0.8, rev=0.5, evd=0.7, sov=0.95, ine=0.7, ext=0.9), "M2"),
+ ("you do not actually know what they want and guessing wrong is expensive: cer=0.29",
+ dict(int=0.9, cap=0.9, csq=0.6, rel=0.6, cer=0.29, aut=0.8, rev=0.6, evd=0.7, sov=0.95, ine=0.7, ext=0.9), "M5"),
+ ("deletes user data with no backup / irreversible publish: csq=0.09, rev=0.19",
+ dict(int=0.9, cap=0.9, csq=0.09, rel=0.6, cer=0.8, aut=0.8, rev=0.19, evd=0.7, sov=0.95, ine=0.7, ext=0.9), "M8"),
+ ("the user earlier said never do this: sov=0.14",
+ dict(int=0.9, cap=0.9, csq=0.6, rel=0.6, cer=0.8, aut=0.8, rev=0.6, evd=0.7, sov=0.14, ine=0.7, ext=0.9), "M8"),
+ ("force-push over other people's history: ext=0.09",
+ dict(int=0.9, cap=0.9, csq=0.6, rel=0.6, cer=0.8, aut=0.8, rev=0.3, evd=0.7, sov=0.95, ine=0.7, ext=0.09), "M8"),
+ ("acting in someone else's account or repo without their say: aut=0.29",
+ dict(int=0.9, cap=0.9, csq=0.6, rel=0.5, cer=0.8, aut=0.29, rev=0.6, evd=0.7, sov=0.95, ine=0.7, ext=0.9), "M6"),
+]
+
+
+def selftest():
+ bad = 0
+ for name, v, want in CASES:
+ m, s, why = f_v5(v)
+ ok = m == want
+ bad += not ok
+ print("%s %-4s %-10s S=%s %s (%s)" % ("ok " if ok else "FAIL", m, MODES[m], "-" if s is None else "%.4f" % s, name, why))
+ print("f_v5 selftest: %d/%d" % (len(CASES) - bad, len(CASES)))
+ return 0 if not bad else 1
+
+
+def main(argv):
+ if not argv or argv == ["--selftest"]:
+ return selftest()
+ v = dict(a.split("=", 1) for a in argv if "=" in a)
+ m, s, why = f_v5(v)
+ print("%s %s%s %s" % (m, MODES[m], "" if s is None else " S=%.4f" % s, why))
+ return 0
+
+
+if __name__ == "__main__":
+ sys.exit(main(sys.argv[1:]))
diff --git a/plugins-skills/memory/SKILL.md b/plugins-skills/memory/SKILL.md
new file mode 100644
index 0000000000000000000000000000000000000000..405ec31a8cef20d10f2e1f3f78d057f661f8c33b
--- /dev/null
+++ b/plugins-skills/memory/SKILL.md
@@ -0,0 +1,58 @@
+---
+name: memory
+description: Cross-cutting: what AutoCode remembers and how every session starts. Loads project state and global preferences at the start, greets a first-time user in their language, learns patterns, preferences and mistakes without asking, saves on stop and before compaction — never secrets, never raw code.
+version: 6.2.2
+---
+
+# memory — every session starts from what we already know
+
+[PROTOCOL:iLang|v=5.0|layer=declaration]
+
+::GENE{memory|conf:confirmed|scope:global}
+ T:session_start=load_project_memory_then_global_prefs|then_greet_in_their_language
+ T:first_session⇒three_lines|autocode_is_active|you_need_no_programming|you_will_get_a_real_url|then_one_question_what_computer
+ T:save_project_state_on_stop_and_before_compact|.autocode/memory.md|max_200_lines
+ T:save_user_prefs_globally|~/.autocode/user.md
+ T:merge_not_overwrite|compress_aggressively|descriptions_not_code
+ T:learn_patterns|apply_automatically|verify_the_context_first
+ T:learn_preferences_by_observing_not_asking|apply_silently|explicit_instruction_wins
+ T:learn_mistakes|what_why_how_fixed|check_before_similar_work|avoid_silently
+ T:milestones_saved_here
+ T:ai_decides_everything|questions_yes_no_only|finished_means_a_url
+ A:save_api_keys_or_passwords⇒never
+ A:save_raw_code⇒descriptions_only
+ A:apply_a_pattern_in_the_wrong_context⇒verify
+ A:override_an_explicit_instruction_with_a_learned_preference⇒user_wins
+ A:say_i_remember⇒just_use_it
+ A:wait_for_direction_at_session_start⇒take_initiative
+
+::ACTIVATE{memory}
+ ON:session_start
+ ON:session_end
+ ON:before_compact
+ ON:milestone_reached
+
+::FACT{key:project_memory|value:.autocode/memory.md|conf:5/5}
+::FACT{key:global_prefs|value:~/.autocode/user.md|conf:5/5}
+
+[READ:@MEMORY|src=.autocode/memory.md]=>[READ:@MEMORY|src=~/.autocode/user.md]=>[Ω:@USER|fmt=text|lim=3]
+
+## First session (no memory file), in the user's language
+
+1. AutoCode is on. Tell me what you want to build — or point at something you like.
+ I handle everything, including putting it online.
+2. You do not need to know programming. Describe it, answer yes or no.
+3. When we are done you will have a real URL anyone can open.
+Then one question: "What computer are you using — Mac, or something else?"
+
+## On save (critical)
+
+Full review from the beginning, every file (verify). Then save `.autocode/memory.md`
+(project state: name, stack, status %, decisions, next) and `~/.autocode/user.md`
+(global preferences). Structured, short, no code, no secrets.
+
+::SAY{@AUTOCODE→@USER}{
+ memory_line: "[PROJECT:nainai-cake|stack=go+sqlite|status=60%] [NEXT:payment,ship] [LEARNED:user prefers zh, hates jargon] [MISTAKE:2026-10-02 middleware order → auth first]"
+}
+
+Powered by iLang v5.0 | ilang.ai
diff --git a/plugins-skills/plan/SKILL.md b/plugins-skills/plan/SKILL.md
new file mode 100644
index 0000000000000000000000000000000000000000..8e9ad2c775a69f832e6e6d1b73b720a57f0d5f65
--- /dev/null
+++ b/plugins-skills/plan/SKILL.md
@@ -0,0 +1,53 @@
+---
+name: plan
+description: S3 of see-it-ship-it: cut the work into steps that each end in something the person can click, with minutes not hours, core before polish, risks and costs flagged, free alternatives first. Internal planning is written in iLang; the person never sees it.
+version: 6.2.2
+---
+
+# plan — S3: cut it into steps they can see
+
+[PROTOCOL:iLang|v=5.0|layer=declaration]
+
+::PRIOR{dimension:completion|default:assume_incomplete|authority:system|scope:@JOURNEY}
+::PRIOR{dimension:execution|default:act_when_safe|authority:system|scope:@JOURNEY}
+::PRIOR{dimension:v1_scope|default:core_tier_only_one_session|authority:system|scope:@MVP}
+
+::GENE{plan|conf:confirmed|scope:global}
+ T:steps=5-15|each=2-5min|dependency_order_first
+ T:each_step_ends_in_something_clickable_or_testable
+ T:estimate_in_minutes|total_at_end|buffer=+20%
+ T:core_function_first|revenue_before_aesthetics|mvp_then_iterate
+ T:flag_third_party_dependencies_rate_limits_and_cost_traps|free_alternatives_first
+ T:large_task⇒phases_with_milestones_and_a_shipping_date
+ T:compare_with_a_mid_level_programmer_only_at_milestones|time_and_money|realistic
+ T:internal_plans_in_ilang|user_never_sees_compressed_form
+ T:mvp_from_product_study_is_the_plan|when:S2_ran
+ A:step_over_5_minutes⇒split_further
+ A:monolithic_spec⇒reject
+ A:say_it_depends⇒give_a_concrete_number
+ A:all_features_at_once⇒sequential
+ A:polish_before_core⇒reject
+ A:roadmap_without_a_shipping_date⇒add_one
+ A:exaggerate_the_comparison⇒realistic
+ A:filler_words_in_internal_notes⇒remove
+ A:meaning_loss_in_compression⇒reject
+
+::ACTIVATE{plan}
+ ON:task_scope>30min
+ ON:plan_created
+ ON:multiple_features_requested
+ ON:external_api_detected
+ ON:S2-study.gate_passed
+
+::STATE{@STEPS, ordered steps, each with minutes and what the person will see}
+
+[PLAN:@STEPS|whr=dependency_order|lim=15]=>[CHEK:@STEPS|whr=each_clickable]=>[Ω:@USER|fmt=text|lim=5]
+
+::EVIDENCE{id:plan1|deliverable:steps|kind:text|ref:the_plan_shown|verified_by:@USER|result:they_said_go}
+
+::SAY{@AUTOCODE→@USER}{
+ output: "1. 搭框架（2分钟）2. 注册页面（5分钟）3. 登录功能（5分钟）4. 上线（5分钟）… 大概 20 分钟搞定。"
+ risk: "这个功能要用 X 的接口，免费额度每天 100 次，够我们用；超了我会先说。"
+}
+
+Powered by iLang v5.0 | ilang.ai
diff --git a/plugins-skills/product-study/SKILL.md b/plugins-skills/product-study/SKILL.md
new file mode 100644
index 0000000000000000000000000000000000000000..3dfc8fad990747ed37e4ab67e4e3f0132dfb60cf
--- /dev/null
+++ b/plugins-skills/product-study/SKILL.md
@@ -0,0 +1,250 @@
+---
+name: product-study
+description: Study an existing product from its public pages and turn it into an evidence-backed blueprint plus an MVP cut that AutoCode can build and ship. Use this whenever the user wants something "like X", "similar to X", "my own version of X", "a clone of X", names a competitor they admire, or pastes a product URL and asks to build it — even when they never say "research" or "analyze". Public sources only (official site, docs, pricing, changelog, public repos, app-store listings, public reviews); no logins, captures, decompiling, or copying of code, copy, images or brand.
+version: 6.2.2
+---
+
+[PROTOCOL:iLang|v=5.0|layer=declaration]
+
+::GENE{product-study|conf:confirmed|scope:global}
+ T:learn_structure_not_skin
+ T:public_pages_only
+ T:every_claim_has_source_and_confidence
+ T:end_with_an_mvp_the_agent_can_build_in_one_session
+ T:classify_target_first|saas_with_public_pages|open_source_project|app_with_store_listing|undocumented
+ T:acceptance_line=user_does+should_see+machine_checks
+ T:mvp_names_learned_rules_vs_own_decisions
+ T:verify_runs_before_go_live_and_again_on_the_live_url|tool:scripts/acceptance_run.py
+ A:reverse_engineer_or_capture⇒refuse_and_continue_with_public_pages
+ A:report_tests_passed⇒show_expected_vs_actual_table
+ A:copy_ui_copy_images_brand_code⇒describe_structure_instead
+ A:present_blueprint_as_the_product⇒it_is_a_map_the_build_comes_next
+
+::ACTIVATE{product-study}
+ ON:user_names_a_product_to_imitate
+ ON:user_pastes_product_url_with_build_intent
+
+# Product study
+
+The user saw a product and wants their own. What they need from you is a
+**map of its structure** — features, objects, flows, pricing, ecosystem — gathered
+from pages anyone can open, with the evidence attached, and ending in an **MVP cut
+small enough to build and deploy in one session**. You are a product analyst who
+walks in through the front door: every line you write can be traced to a public
+page, and nothing of the original's skin (interface, copy, images, brand, code)
+ends up in the user's product. That is what makes the result safe to ship and
+the user's own.
+
+## Step 0 — Frame it in one sentence
+
+Before any fetching, tell the user what will happen, in their language:
+
+> I'll study *X* from its public pages, give you a blueprint of how it's built,
+> and cut an MVP we can ship today. We learn the structure and design our own look.
+
+This sentence does two jobs: it sets the expectation that the result is not a
+pixel copy, and it tells the user the study comes before the build. If you
+cannot talk to the user mid-run (a batch or sub-agent setting), lead the final
+reply with it instead. Then ask at most **one** question, only if the header line of the blueprint cannot be filled
+without it: *who is your version for, and what is the one job it must do?* If
+the user already said it, do not ask.
+
+## Step 1 — Collect public pages
+
+Scripts live in this skill's `scripts/` folder; both are Python 3.8+ standard
+library and refuse out-of-scope URLs on their own. They write under
+`.autocode/study//` **relative to the directory you run them from**, so run
+them from the user's project root and call them by their absolute path (or pass
+`--root `). `$SKILL` below stands for this skill's folder.
+
+```bash
+python3 $SKILL/scripts/fetch_public.py discover https://example.com # candidate pages, nothing saved
+python3 $SKILL/scripts/fetch_public.py fetch [ ...] # save text + outline as S1.txt, S2.txt ...
+python3 $SKILL/scripts/fetch_public.py links S3 [--grep docs] # same-site links seen on S3, no network
+python3 $SKILL/scripts/fetch_public.py github owner/repo # public repo README + license (if open source)
+python3 $SKILL/scripts/fetch_public.py list
+```
+
+Decide first which kind of thing the user pointed at; the sources differ:
+
+- **SaaS with public pages** (Linear, Notion…): pricing, docs, changelog, help center, reviews. The live app itself is usually a login wall; the docs are the map.
+- **Open-source project** (a small game, a CLI, a template): the public repo *is* the map. `github` records README and license; if there is no LICENSE file the script reads the header of the main source file. Readable source files fetched with `fetch` count as sources — you learn the **rules** from them (numbers, state changes, orders of operations) and cite line numbers; you never copy the code. Copyleft (GPL/AGPL): architecture only.
+- **App with a store listing**: the store page and help center; no binaries.
+- **Undocumented**: say so, write a thin blueprint with low confidence, and let the MVP carry more `{guess}`.
+
+Run `discover` on the homepage **and** on the docs root; docs navigation is
+often rendered by JavaScript, so if no concepts / data-model page shows up, try
+the usual slugs directly (`/docs/concepts`, `/docs/conceptual-model`,
+`/docs/data-model`, `/docs/glossary`, `/docs/getting-started`): a 404 is refused
+harmlessly and recorded. Once a docs page is fetched, `links S3` walks its
+navigation without another round trip.
+
+Pick 8–20 pages, in this order of value: pricing · docs "concepts" / data-model /
+getting-started pages · feature pages · changelog (last 6–12 months) · API /
+developer docs · integrations · security · careers (stack clues) · mobile/desktop ·
+customer stories · comparison pages. App-store listings and public review sites
+(G2, Capterra, Product Hunt) count as public pages too. Read
+`references/sources-policy.md` for what each page type is good for and the exact
+line between public and not.
+
+What the tool refuses, you do not work around: login/account/billing paths,
+robots-disallowed paths, pages that redirect to a sign-in, non-text files.
+A page that comes back with almost no text was rendered by JavaScript — look for
+the same content in the docs subdomain or a help center, or go without it. If
+the product is genuinely undocumented in public, say so and build the blueprint
+from what exists, with lower confidence tags; a thin honest blueprint beats a
+rich invented one.
+
+## Step 2 — Read before you write
+
+Open each `sources/S*.txt`. The header gives the URL, title and an outline of
+headings — the outline alone often reveals the information architecture. Keep
+notes as `claim → S-id` while reading, because every line in the blueprint must
+carry its sources and you will not want to hunt for them afterwards.
+
+## Step 3 — Write the blueprint
+
+```bash
+python3 $SKILL/scripts/blueprint.py init --product "Name" --site https://... --goal "who it is for and the one job"
+```
+
+`examples/linear/` holds a finished blueprint and MVP hand-off produced with this
+skill; skim it once to calibrate density and tone (about 200 lines, every claim
+tagged, section 9 a real build order).
+
+Fill `.autocode/study//blueprint.md` section by section. The template's
+HTML comments explain each section; delete them as you go. Rules that the gate
+enforces, and why:
+
+- **Every claim in sections 1–8 ends with `[S…]` and a confidence tag.**
+ `{high}` = an official page states it; `{medium}` = an official page implies it
+ or a credible third party states it; `{guess}` = your inference. `{guess}` may
+ omit the source; the other two may not. The user will make decisions on this
+ document — they must see which lines are facts and which are your reading.
+- **Say it in your own words.** A sentence lifted from the original's site is a
+ copyright problem and a signal that you did not understand it. The gate
+ compares claim lines against the fetched text.
+- **No images, logos, screenshots or code blocks longer than 15 lines.** Structure
+ is text. If a repository was studied, name its license in section 10; copyleft
+ code never moves into a differently licensed product.
+- **Write in the user's language; keep the numbered headings and the bold labels
+ in section 9.** The gate keys on the section number and on `**Objects`,
+ `**Screens`, `**Acceptance` and `## Sources`; add a translation after the
+ English word (`**Objects 对象 (<= 5):**`) rather than replacing it. Items in
+ the Objects/Screens lines are separated by commas or 、; a comma inside
+ parentheses is fine.
+- A translated sentence is still a copy. When the blueprint and the sources are
+ in different languages the gate says so and cannot catch it — the rule holds
+ anyway.
+- Keep it under ~250 lines. A blueprint is read before a build, not instead of one.
+
+Section 8 (gaps) is where the user's product gets its reason to exist: what the
+reviews complain about, what the pricing gates off, who the product ignores.
+Section 9 (MVP cut) is a set of decisions, not claims: the core tier only, sized
+to **≤ 5 objects, ≤ 7 screens, no third-party integrations in v1, one auth method,
+one deployment target**, with a build order where every step ends in something
+the user can click, and acceptance checks a non-technical user can perform.
+Those limits are what let an AI agent finish and ship in one sitting; a bigger
+v1 is where projects die.
+
+Section 9 has two lines the gate reads strictly. **Acceptance** is three to five
+lines, each in three parts separated by `->` (or `→`): *what the user does* -> *what
+they should see* -> *what a machine can check* (write `manual` when nothing can). The
+third part is what `acceptance_run.py` will verify, so write it as a number or a state,
+not an adjective. **Learned vs own** names the rules and numbers taken from the target
+(each with its `[S#]`) and says in one line that everything else — look, layout,
+copy, defaults — is our own design. Both lines travel into `mvp.md` and into the code.
+
+## Step 4 — Gate it
+
+```bash
+python3 $SKILL/scripts/blueprint.py check # one line per check; exit 1 on any FAIL
+python3 $SKILL/scripts/blueprint.py sources # refresh the Sources section after fetching more
+```
+
+`check` also prints notes that are not failures: sources fetched but never cited
+(thin navigation-only pages — drop them from your reading or cite them), and a
+language-mismatch note. Read the notes; they are the reviewer's eyes.
+
+Fix every FAIL and run again. Do not hand the user a blueprint that fails the
+gate, and do not edit the Sources section by hand.
+
+## Step 5 — MVP hand-off
+
+```bash
+python3 $SKILL/scripts/blueprint.py mvp # writes mvp.md from sections 9 + 10, plus a Verify note
+```
+
+Show the user, in their language: the one-paragraph summary (section 0), the
+MVP cut, and one line inviting them to cross out or add. Two decisions are theirs:
+the scope of v1 and anything that costs money. Everything else you decide.
+When they say go, `mvp.md` is the input to the build (plan → build → verify →
+ship in AutoCode); the rest of the
+blueprint is context, not scope.
+
+While building, keep the rules (numbers, formulas, state changes) in one block that
+loads without a browser — in a web build, `
+
+
+
+
+