claude-code-docs

name	claude-code-docs
description	Look up authoritative, up-to-date Claude Code, Agent SDK, hooks, MCP, skills, plugins, settings, subagents, and admin documentation. Use when the user asks how a Claude Code feature works, what a config field does, how to set up hooks/MCP/skills/subagents/plugins, when troubleshooting a Claude Code error or unexpected behavior, or when you need to cite current official docs rather than rely on training-cutoff knowledge.
argument-hint	["topic or doc title"]
allowed-tools	WebFetch

Claude Code Docs

Lazy-loads the official Claude Code documentation by reading the index at https://code.claude.com/docs/llms.txt, picking the most relevant page(s), and fetching them on demand. Always prefer this skill over recalling docs from memory — the docs change weekly.

If the user passed an argument ($ARGUMENTS), treat it as the topic to look up. Otherwise infer the topic from the conversation.

When to use this skill

Use it whenever the question is about Claude Code itself or anything in its docs scope:

• Claude Code CLI: slash commands, settings, permissions, hooks, MCP, plugins, subagents, skills, output styles, memory/CLAUDE.md
• Claude Agent SDK (Python or TypeScript): API surface, sessions, streaming, hooks, custom tools, hosting, permissions, observability
• IDE/Desktop/Web integrations, Bedrock/Vertex/Foundry deployment, GitHub/GitLab CI, admin setup, troubleshooting

If the question is about the Anthropic API / Anthropic SDK (not Claude Code), use the claude-api skill instead.

Procedure

1. Read the index

WebFetch url=https://code.claude.com/docs/llms.txt
        prompt="Return the raw markdown. I need every `- [Title](URL): description` line unmodified."

The index is a flat list of ~150 entries, each in the form - [Title](https://code.claude.com/docs/en/<slug>.md): description. URLs end in .md — the targets are raw Markdown, not HTML.

2. Pick the right page(s)

Match the user's question against the description (text after the colon), not just the title. Then:

• Pick 1–3 pages per batch, not more. The index is for triage, not bulk loading.
• One specific feature ("how do hooks work?") → one page.
• Cross-concept question ("how do skills relate to subagents?") → fetch each relevant page.
• Nothing in the index obviously matches → say so. Do not guess a URL.

3. Fetch the batch

For each chosen URL:

WebFetch url=<URL from index>
        prompt="<a question that captures what the user actually needs, not 'summarize this page'>"

4. Evaluate, then loop or answer

After each batch, judge whether the fetched pages actually answer the user's question:

• Enough → answer, grounded in the fetched content. Cite the doc page (title + URL) when stating non-obvious facts so the user can verify.
• Not enough (the answer lives on a page you haven't read, or a fetched page pointed to another) → go back to step 2, pick the next 1–3 pages from the index, and fetch again.
• Keep looping until you can answer, up to a default cap of 9 pages total across all batches.
• Still not enough at 9 pages → stop. Tell the user honestly what you've read, what's still missing, and ask whether they want you to keep reading more pages. Don't silently blow past the cap or pad the answer with guesses.

Rules

• Never invent a doc URL. If a page isn't in the index, it does not exist — say so instead of fabricating a slug.
• Don't skip step 1, even if you think you remember the right URL. Doc slugs get renamed; the index is the source of truth.
• Loop in small batches, cap at 9 pages. Fetch 1–3, check if that's enough, fetch more only if it isn't. If 9 pages still don't answer it, ask the user before reading more — don't grind through the whole index or fabricate the gap.
• Stay in scope. This skill covers code.claude.com/docs/* only. For Anthropic API use claude-api.
• Pass through what the docs say. Don't merge aggressively with prior knowledge — the user wants current authoritative behavior, not a synthesis.