Marketplace for trusted AI agent skills

Skill Detail

Answer Anthropic documentation questions from a versioned local archive

Use Claude Code to investigate one API, Claude Code or MCP question against a local documentation snapshot with paths, provenance and freshness limits.

Library & API ReferenceClaude Code
Library & API Reference Claude Code Security Reviewed
⭐ 154 GitHub stars
COPY SKILL INSTRUCTIONS (OPTIONAL)
npx skills add agentskillexchange/skills --skill answer-anthropic-documentation-questions-from-a-versioned-local-archive Copy
Uses the third-party skills CLI, not an ASE-owned installer. Check your agent’s compatibility. This copies instructions; complete the upstream tool setup below separately.
View source Documentation
At a glance
Tools required
Git; Claude Code; local text search; uv only for optional refreshes
Install & setup
git clone https://github.com/thevibeworks/claude-code-docs; cd claude-code-docs. With Claude Code installed and configured, run claude "show me the MCP spec for tool annotations" or provide your own bounded documentation question. Ask for the checkout commit and exact supporting file paths. No fetcher run is needed for an existing snapshot; optional refresh commands and credentials are documented upstream.
Author
thevibeworks
Publisher
Community
Last updated
Oct 17, 2026
Quick brief

Invoke this workflow when a Claude Code user needs to answer a bounded Anthropic API, Claude Code or MCP question from a locally searchable, versioned documentation snapshot. Supply the question, relevant product or protocol version, and archive checkout. Return an answer with repository commit, exact file paths, supporting passages, original source links where verified, and explicit freshness limits.

How it works

What this skill actually does

Clone the upstream Claude Code Docs repository and start Claude Code inside it. Its README documents asking questions with the claude command, and CLAUDE.md maps topics to local content paths. Record git rev-parse HEAD, search the relevant section, and read matching passages before answering. Use sources.json to establish source provenance; do not guess the original URL when the mapping is unclear. Distinguish archive evidence from a current upstream guarantee.

The README documents optional refreshes through uv run scripts/fetcher.py –section mcp and other named sections. Refresh a dedicated archive checkout only when the user requests current data and inspect the resulting diff before relying on it. A GitHub-source refresh requires a token; local research of an existing checkout does not. Fetching can update or reap archive files, so leave the application’s working tree separate.

Invoke this instead of ordinary hosted documentation browsing when offline access, a reproducible snapshot or local comparison matters. The boundary is one evidence-backed documentation investigation, excluding generic chat, installing the mirrored skill collection, autonomous source additions or maintaining the archive’s scheduled automation. The engineering/blog archive is frozen as of July 2026; do not present it as refreshed four times daily. Counts and freshness vary by section. This is an unofficial community mirror, not Anthropic’s official documentation. Claude Code is the supported framework because the upstream CLI examples and repository instructions make it central to invocation. Metadata review does not establish archive completeness, answer accuracy, runtime behavior or security.