本文へ移動
cccskills
無料GitHub で公開

claude-docs

Search and read locally-stored Claude documentation covering Claude Code CLI, Claude API (Messages, tool use, vision, streaming, batch), Agent SDK (Python and TypeScript), prompt engineering, and all Anthropic platform docs. Use this skill whenever the user asks about Claude Code features (hooks, MCP servers, skills, plugins, settings, permissions, keybindings, sub-agents), the Anthropic API or any of its SDKs (Python, TypeScript, Go, Java), the Agent SDK (sessions, hooks, custom tools, MCP), model capabilities (context windows, extended thinking, pricing, rate limits, vision), prompt engineering best practices, or troubleshooting any Claude-related error. This skill provides instant access to official documentation files without web searches — always prefer it over web lookups for Claude and Anthropic topics.

インストール方法を見る

含まれるファイル(7)

  • SKILL.md7.0 KB
  • examples/cross-context.md2.4 KB
  • examples/direct-lookup.md1.8 KB
  • examples/semantic-search.md2.0 KB
  • manifest-reference.md2.6 KB
  • scripts/content-search.sh3.5 KB
  • scripts/fuzzy-search.sh3.2 KB

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

Claude Documentation Search Skill

Claude's official documentation is indexed locally. The clone at ~/.claude-code-docs/ holds only metadata:

  • paths_manifest.json — every page's filename, id, title, category, and source URL
  • search_index.json — per-page titles, headings, and stemmed term counts

The actual .md pages are cached at ~/.claude-code-docs/cache/ (override: $CLAUDE_DOCS_CACHE_DIR). A background sync keeps the cache current; any page not yet cached is fetched on demand — see Reading a doc below.

When to Use This Skill

Activate when the user asks about:

  • Claude Code features: hooks, skills, MCP, plugins, settings, slash commands, sub-agents
  • Claude API: messages, tool use, streaming, batch processing
  • Agent SDK: Python/TypeScript SDK, sessions, custom tools, subagents
  • Prompt engineering: best practices, system prompts, chain of thought
  • Any topic covered by platform.claude.com or code.claude.com

Search Strategy

The search scripts read the manifest + index, so they see every page whether or not it is cached yet. Prefer them over globbing the cache.

1. Content search (default — questions and topics)

bash ~/.claude-code-docs/plugin/skills/claude-docs/scripts/content-search.sh "<keyword1>" "<keyword2>"

Output is filename<TAB>title<TAB>score, best first. Keyword extraction: strip filler, keep domain terms — "how do I configure streaming" → streaming configure; "difference between hooks and MCP" → hooks mcp. Take the top 3-5 filenames and read them (next section).

2. Fuzzy search (approximate name)

User says "that caching doc", "something about checkpoint":

bash ~/.claude-code-docs/plugin/skills/claude-docs/scripts/fuzzy-search.sh "<query>"

Output is ranked filenames. Read the top match.

3. Direct manifest lookup (exact category/topic)

To list pages in a category or matching a filename fragment, query the manifest:

jq -r '.pages[] | select(.filename | test("<fragment>")) | .filename' ~/.claude-code-docs/paths_manifest.json
jq -r '.pages[] | select(.category=="claude_code") | .filename' ~/.claude-code-docs/paths_manifest.json

Reading a doc (cache-miss rule)

A search returns a filename (e.g. claude-code__hooks.md). The file is at ~/.claude-code-docs/cache/<filename>.

  1. Read ~/.claude-code-docs/cache/<filename>.
  2. If it is not there (not fetched yet), fetch it first, then read:
    ~/.claude-code-docs/plugin/scripts/fetch-docs.sh get "<filename>"
    
    then Read ~/.claude-code-docs/cache/<filename>.
  3. If the fetch fails (offline), the script prints the canonical source URL on stderr — fall back to WebFetch on that URL.

To save context, prefer previewing large pages before reading them. Pull a page's structure from the index (title + headings) first — often the headings alone answer the question and you skip loading a multi-KB body:

jq -r '.pages[] | select(.filename=="<filename>") | .title, (.headings[]|"  "+.text)' ~/.claude-code-docs/search_index.json

Synthesis Rules

Same Product Context → SYNTHESIZE

When all matching docs share one product (all Claude Code, all Agent SDK, ...): read them all silently, extract relevant sections, present one unified answer, cite sources.

Different Product Contexts → ASK

When matches span products (CLI + API + Agent SDK), ask which the user means. Labels (see manifest-reference.md) map from category:

categorySay to user
claude_codeClaude Code CLI
agent_sdkClaude Agent SDK
api_referenceClaude API
core_documentationClaude Documentation
agents_and_toolsAgents & Tools
about_claudeAbout Claude
get_startedGetting Started
test_and_evaluateTesting & Evaluation
release_notesRelease Notes
resourcesResources
prompt_libraryPrompt Library

After selection → read all docs in that context and synthesize.

SDK Language Disambiguation

When the user names a language, narrow API/SDK results (e.g. agent-sdk__python, agent-sdk__typescript, api__go, api__java). If none named but the query is about SDK methods, present Python first and note TypeScript equivalents exist.

URL Generation

Always cite source links. Get the exact URL from the manifest — never reconstruct it from the filename (filenames are lossy; the manifest stores the verbatim URL):

jq -r '.pages[] | select(.filename=="<filename>") | .url' ~/.claude-code-docs/paths_manifest.json

Example: claude-code__hooks.md → https://code.claude.com/docs/en/hooks.

Common Synonyms

When a search returns too few results, retry with a known synonym:

User saysSearch for
"function calling""tool use", "tool-use"
"system instructions""system prompt"
"JSON mode""structured outputs"
"thinking""extended thinking", "adaptive thinking"
"caching""prompt caching", "prompt-caching"
"files API""files", "pdf support"
"sub-agents""subagents", "sub-agents"
"environment variables""settings", "configuration"
"CI/CD""github actions", "gitlab"

No Results

If all strategies return nothing: try synonyms; try broader/narrower keywords; suggest /docs -t to check freshness; note the topic may not be in the index.

Post-Response Course Prompt

After presenting documentation, always append this block at the bottom:


Want to go deeper? I can create an interactive course on this topic — visual explanations, animated diagrams, and hands-on quizzes in a single HTML file you can open in your browser. Just say "create a course on [topic]" or run /docs --course [topic].

Replace [topic] with the actual topic the user asked about.

Reference Files

  • manifest-reference.md — Category-to-label mapping (single source of truth)
  • examples/direct-lookup.md — Example: topic → search → synthesize
  • examples/semantic-search.md — Example: question → content-search.sh → synthesize
  • examples/cross-context.md — Example: ambiguous → ask context → synthesize

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

Generate a self-contained HTML changelog report showing recent documentation changes with interactive course generation buttons. Use this skill when the user asks for a docs changelog, documentation report, what changed recently, recent doc updates as a report, or runs `/docs --report`. Discovers changes via git history, categorizes them, summarizes what's new, and produces a stunning Obsidian & Amber themed HTML report where each entry has a "Create Course" button that copies the course command to clipboard.

日本語の概要は準備中です。原文の説明を表示しています。

costiash/claude-code-docs542026年10月10日 更新

Generate an interactive, self-contained HTML course on any Claude documentation topic. Use this skill when the user wants to create an interactive course, tutorial, or deep-dive walkthrough about a Claude feature, API concept, SDK pattern, or prompt engineering technique. Triggers on: "create a course about", "interactive tutorial for", "teach me about X interactively", "make a course on", "I'd like a course", "/docs --course", "/docs course", or when the user accepts the post-docs course prompt. Produces a stunning single-page HTML file with scroll-based navigation, animated visualizations, quizzes, and code translations drawn from official Claude documentation.

日本語の概要は準備中です。原文の説明を表示しています。

costiash/claude-code-docs542026年10月10日 更新

Check the health and freshness of locally-stored Claude documentation. Use this skill when the user asks about documentation health, broken links, stale docs, freshness checks, or wants to validate that their local install is up-to-date and all URLs are reachable. Triggers on: "are my docs current", "check doc health", "validate documentation", "broken links", "stale docs".

日本語の概要は準備中です。原文の説明を表示しています。

costiash/claude-code-docs542026年10月10日 更新

docs

無料

Router for the /docs command. Look up Claude documentation, ask plain-English questions, generate interactive courses and HTML changelog reports, check freshness, and sync the local cache. Dispatches to the claude-docs, claude-docs-validate, claude-docs-course, and claude-docs-changelog skills.

日本語の概要は準備中です。原文の説明を表示しています。

costiash/claude-code-docs542026年10月10日 更新

costiash のスキルをすべて見る

このスキルの問題を報告する