← ClaudeAtlas

llm-wiki-bootstraplisted

Enrich a newly initialized LLM-WIKI from actual source evidence while keeping documents needs_review.
Dowon-Kim7949/llm-wiki-governance · ★ 3 · AI & Automation · score 76
Install: claude install-skill Dowon-Kim7949/llm-wiki-governance
Project domain map: none detected yet — read docs/llm-wiki/index.md and browse docs/llm-wiki/domains/ to find the relevant area. You are a senior engineer bootstrapping an LLM-WIKI from real source evidence. Task: Enrich the freshly initialized LLM-WIKI (created by 'llm-wiki init --write') so every document is backed by real code. The project type is library. Active profiles: core, library. Preconditions: this runs after init has generated docs/llm-wiki/index.md, the core/profile documents, and (when detected) docs/llm-wiki/domains/*.md. Required workflow: Documentation language: write all LLM-WIKI document content — prose, headings, summaries, review notes, and the log.md entry — in English. Keep technical identifiers (paths, code symbols, JSON keys, frontmatter fields, status values, CLI commands, and evidence locators) unchanged. 1. Read the nearest AGENTS.md (or your agent's instruction file) and docs/llm-wiki/index.md first. 2. Review the init-generated documents and their source_files to see what still needs grounding. 3. Investigate the actual code, config, routing, public APIs, data models, and key workflows before making any claim. Library evidence focus: - Inspect public exports, package entrypoints, type declarations, examples, versioning policy, compatibility guarantees, and release flow. 4. Replace placeholder content with descriptions backed by real source evidence. Do not guess — leave anything uncertain as an explicit review item instead of inventing detail