From anthropic-docs
Routes Claude Code questions to topic-specific reference files covering settings, hooks, slash commands, MCP, plugins, and CLI internals.
How this skill is triggered — by the user, by Claude, or both
Slash command
/anthropic-docs:claude-codeThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
| Field | Value |
CHANGELOG.mdREADME.mdSKILL-cli.mdSKILL-hooks.mdSKILL-known-issues.mdSKILL-mcp.mdSKILL-plugins.mdSKILL-settings.mdSKILL-slash-commands.mdconfig.jsondocs-snapshot/MANIFEST.jsonreports/2026-05-19.mdreports/2026-05-20.mdreports/2026-05-21.mdreports/2026-05-22.mdreports/2026-05-23.mdreports/2026-05-24.mdreports/2026-05-25.mdreports/2026-05-26.mdreports/2026-05-27.md| Field | Value |
|---|---|
| Claude Code version | v2.1.160 (pinned in CI; updated by pipeline) |
| Source docs | code.claude.com/docs |
| GitHub | anthropics/claude-code |
| npm | @anthropic-ai/claude-code |
This skill is auto-updated daily. A matrix pipeline reads the upstream docs and rewrites the per-surface files below. Section structure is stable; content drifts to track upstream.
Read the surface file(s) matching the user's question. Each surface file holds its complete reference for that one topic and cross-links (via markdown links) to siblings when a fact spans surfaces — so you only need to Read the matching file(s), not all of them.
| Surface file | Read when the user asks about… |
|---|---|
SKILL-settings.md | settings.json, settings.local.json, project vs user vs local scope, settings keys / defaults / types |
SKILL-hooks.md | hook events (PreToolUse, PostToolUse, Stop, SubagentStop, Notification, UserPromptSubmit, PreCompact, SessionStart, SessionEnd), hook input/output JSON shapes, matchers, blocking vs non-blocking |
SKILL-slash-commands.md | slash command authoring, frontmatter schema, argument syntax, $ARGUMENTS, command discovery paths |
SKILL-mcp.md | .mcp.json schema, MCP transports (stdio / http / sse), server config, MCP tool naming, mcp__ prefix |
SKILL-plugins.md | plugin manifest (.claude-plugin/plugin.json), marketplaces (marketplace.json), plugin install / scope, plugin commands / agents / skills / hooks, CLI hint protocol (<claude-code-hint />) for recommending plugins from third-party CLIs |
SKILL-cli.md | CLI flags, subcommands, env vars (ANTHROPIC_* / CLAUDE_*), permission modes (default / acceptEdits / plan / bypassPermissions), ~/.claude/ directory layout, IDE integrations, session management (resume, naming, branching, picker, transcript storage), sandbox environments and isolation approaches (sandboxed Bash tool, sandbox runtime, dev containers, VMs) |
SKILL-known-issues.md | a user reports a bug, asks about a workaround, mentions an error message, or asks "why does X not work" |
.mcp.json, settings.json, plugin.json), read the matching surface file.SKILL-known-issues.md first.SKILL-cli.md — it's the broadest catch-all surface.Path-scoped correction rules live in rules/:
| Rule file | Triggers on edits to |
|---|---|
rules/settings.md | **/.claude/settings.json, **/.claude/settings.local.json |
rules/mcp.md | **/.mcp.json |
rules/plugins.md | **/.claude-plugin/plugin.json, **/marketplace.json |
rules/hooks.md | **/.claude/hooks/** |
rules/skills-agents-commands.md | **/.claude/skills/**/SKILL.md, **/.claude/agents/**, **/.claude/commands/** |
This skill is auto-updated daily by a maintainer-run pipeline. If you spot a bug in this content, file an issue at xiaolai/anthropic-docs — SKILL.md fixes flow back through the next research run, not via PRs.
npx claudepluginhub xiaolai/claude-plugin-marketplace --plugin anthropic-docsProvides official documentation for Claude Code CLI, plugins, hooks, MCP servers, skills, configuration, and features. Use for setup, development, troubleshooting, and integrations.
Answers questions about Claude Code features including setup, best practices, automation, models, plugins, MCP, and configuration by reading reference documentation.
Claude Code extensibility and configuration reference: plugins, hooks, skills, subagents, MCP servers, output styles, memory, settings, and model configuration. Invoke whenever Claude Code itself is the subject — questions, configuration, building extensions, debugging, or understanding internals.