← ClaudeAtlas

speckit-docguard-generatelisted

Reverse-engineer canonical documentation from existing codebase
jaroslawherod/micold-ai-ide · ★ 1 · Testing & QA · score 73
Install: claude install-skill jaroslawherod/micold-ai-ide
# Docguard Generate Skill # DocGuard Generate Scans your codebase (JS/TS, Python, Rust, Go, Java/Kotlin, Ruby, PHP, C# — polyglot/monorepo-aware) and generates the canonical documentation memory: ARCHITECTURE.md, DATA-MODEL.md, TEST-SPEC.md, SECURITY.md, ENVIRONMENT.md, API-REFERENCE.md, SCREENS.md. Two modes: - **`--plan`** (AI-powered, recommended) — emits a structured agent task manifest + writes the code-truth skeleton inside `<!-- docguard:section -->` markers. The AI agent then writes the prose grounded in scanned facts. Human prose is preserved. - **default** — purely deterministic generation: writes templated docs with TODO placeholders. Use when no AI agent is available. ## User Input $ARGUMENTS ## Steps 1. **Preview the plan** — what code-truth facts were captured + what the agent will write: ```bash npx --yes docguard-cli@latest generate --plan $ARGUMENTS ``` 2. **Scaffold the skeleton docs** (marked sections filled with code-truth, prose sections as agent-task placeholders): ```bash npx --yes docguard-cli@latest generate --plan --write $ARGUMENTS ``` 3. **Or get the machine-readable manifest** to drive an agent: ```bash npx --yes docguard-cli@latest generate --plan --format json $ARGUMENTS ``` 4. **Fallback** (no AI, deterministic generation): ```bash npx --yes docguard-cli@latest generate $ARGUMENTS ``` 2. Review the generated docs in `docs-canonical/`. Each document includes: - Structured sections based on industry standards - Data extracte