document

/document — Auto-Generate and Maintain Documentation

When to use

When the user says "document", "docs", "docstring", "README", "API docs", "architecture", "changelog", or wants to create, update, or fill gaps in project documentation.

Complements /explore (understanding code) by converting that understanding into persistent, readable documentation.

Steps

1. MEMORY CHECK

• search_knowledge("documentation") — find prior documentation decisions and patterns
• search_knowledge("[project name] conventions") — find project-specific style rules
• search_knowledge("type:decision area:docs") — find past documentation choices
• If documentation patterns exist, follow them for consistency

2. DETECT GAPS

Scan the project for missing or outdated documentation:

Check for:

• README.md — Does it exist? Is it up to date? Does it cover setup, usage, architecture?
• Docstrings — Scan public functions/classes for missing docstrings
  • Grep for def and class definitions, check if preceded by docstrings
• API documentation — Are endpoints documented? Are request/response schemas described?
• CHANGELOG — Does one exist? Is it current with recent changes?
• Configuration docs — Are environment variables, config files, and settings documented?
• Architecture docs — Is the high-level structure documented anywhere?

Present the gap analysis to the user:

Documentation gaps found:
  [x] README.md — exists but missing architecture section
  [ ] API docs — no endpoint documentation found
  [x] Docstrings — 12/30 public functions have docstrings (40%)
  [ ] CHANGELOG — not found
  [x] Config — .env.example exists with comments

Ask the user which gaps to fill.

3. ANALYZE CODE

For the selected documentation targets, extract information from the code:

For README/setup docs:

• Find entry points (main.py, index.js, Makefile, package.json scripts)
• Identify dependencies (requirements.txt, package.json, go.mod)
• Find configuration (env vars, config files, CLI arguments)
• Check for Docker/container setup