claude-code-reference

name	claude-code-reference
description	Claude Code 官方文档参考技能。When working with Claude Code features, configuration, hooks, skills, MCP, permissions, settings.json, CLI commands, slash commands, sub-agents, Agent SDK, or any other Claude Code internals — either when users ask OR when the Agent needs to know how something works. Fetch from https://code.claude.com/docs/ as the authoritative source instead of relying on training data. Make sure to use this skill whenever the Agent encounters a Claude Code feature it doesn't fully understand, needs to configure settings.json, create hooks/scripts, use MCP servers, work with skills, or needs accurate information about Claude Code commands, permissions, or APIs.
compatibility	WebFetch, Agent

Claude Code Reference Skill

Core Principle

Always prefer fetching from official documentation over relying on training data. When working with Claude Code features, use the official docs at https://code.claude.com/docs/ as the authoritative source. This applies both when users ask AND when the Agent needs accurate information about how Claude Code works internally.

When to Use This Skill

Trigger in all these contexts:

• User asks about Claude Code features, configuration, or commands
• Agent needs to implement hooks, skills, MCP servers, or plugins
• Agent needs to configure settings.json correctly
• Agent encounters an unfamiliar Claude Code feature or command
• Agent needs to write Claude Code scripts (hooks, slash commands, plugins)
• Agent needs to understand permissions, permission modes, or sandboxing
• Agent needs to use sub-agents, agent teams, or worktrees
• Agent needs to integrate with GitHub Actions, GitLab CI, Slack, or other platforms
• Agent needs to troubleshoot Claude Code issues
• Agent needs accurate CLI reference, environment variables, or tool usage

Official Documentation Index

The full documentation index is at: https://code.claude.com/docs/llms.txt

Key documentation pages (use WebFetch to retrieve):

Getting Started

• Overview: /docs/en/overview.md
• Quickstart: /docs/en/quickstart.md
• Setup: /docs/en/setup.md
• Authentication: /docs/en/authentication.md

Core Configuration

• Settings: /docs/en/settings.md
• Environment variables: /docs/en/env-vars.md
• Model configuration: /docs/en/model-config.md
• CLI reference: /docs/en/cli-reference.md
• Commands: /docs/en/commands.md
• Tools reference: /docs/en/tools-reference.md

Agent Features

• Skills: /docs/en/skills.md
• Hooks guide: /docs/en/hooks-guide.md
• Hooks reference: /docs/en/hooks.md
• MCP: /docs/en/mcp.md
• Plugins: /docs/en/plugins.md
• Plugins reference: /docs/en/plugins-reference.md
• Sub-agents: /docs/en/sub-agents.md

Memory & Context

• Memory: /docs/en/memory.md
• Best practices: /docs/en/best-practices.md
• Common workflows: /docs/en/common-workflows.md
• Context window: /docs/en/context-window.md

Workflow & Automation

• Routines: /docs/en/routines.md
• Scheduled tasks: /docs/en/scheduled-tasks.md
• Agent teams: /docs/en/agent-teams.md
• Worktrees: /docs/en/worktrees.md

UI & Customization

• Interactive mode: /docs/en/interactive-mode.md
• Terminal config: /docs/en/terminal-config.md
• Keybindings: /docs/en/keybindings.md
• Status line: /docs/en/statusline.md
• Output styles: /docs/en/output-styles.md

Permissions & Security

• Permission modes: /docs/en/permission-modes.md
• Configure permissions: /docs/en/permissions.md
• Security: /docs/en/security.md
• Sandboxing: /docs/en/sandboxing.md

Integrations & Platforms

• Desktop app: /docs/en/desktop.md
• VS Code: /docs/en/vs-code.md
• JetBrains: /docs/en/jetbrains.md
• Web: /docs/en/claude-code-on-the-web.md
• GitHub Actions: /docs/en/github-actions.md
• GitLab CI/CD: /docs/en/gitlab-ci-cd.md
• Slack: /docs/en/slack.md
• Third-party integrations: /docs/en/third-party-integrations.md

Agent SDK

• Agent SDK overview: /docs/en/agent-sdk/overview.md
• Skills in SDK: /docs/en/agent-sdk/skills.md
• Hooks in SDK: /docs/en/agent-sdk/hooks.md
• MCP in SDK: /docs/en/agent-sdk/mcp.md
• Permissions in SDK: /docs/en/agent-sdk/permissions.md

Troubleshooting

• Troubleshooting: /docs/en/troubleshooting.md
• Debug your config: /docs/en/debug-your-config.md
• Error reference: /docs/en/errors.md
• Troubleshoot install: /docs/en/troubleshoot-install.md

Observability

• Monitoring usage: /docs/en/monitoring-usage.md
• Costs: /docs/en/costs.md
• Analytics: /docs/en/analytics.md

Workflow

Step 1: Identify the Topic

Map the question to the most relevant documentation page(s). Use the index above for guidance.

Step 2: Fetch from Official Docs

Use WebFetch to retrieve the relevant page(s):

WebFetch(url="https://code.claude.com/docs/en/[topic].md", prompt="Extract all relevant information about [specific topic]")

For unknown topics, fetch the index first:

WebFetch(url="https://code.claude.com/docs/llms.txt", prompt="Find pages related to [topic]")

Step 3: Synthesize and Answer

Combine information from official docs with your own knowledge. When citing specific features, configuration options, or command syntax, always reference the official documentation as the source.

Step 4: Provide Actionable Guidance

Don't just quote docs — provide concrete examples, configuration snippets, or step-by-step instructions based on what the official docs say.

Examples

User asks about hooks: → Fetch /docs/en/hooks.md and /docs/en/hooks-guide.md, explain hook types, configuration, and provide examples.

Agent needs to create a skill: → Fetch /docs/en/skills.md, explain skill structure, SKILL.md format, and how to trigger skills.

Agent needs to configure MCP: → Fetch /docs/en/mcp.md, explain MCP server setup, settings.json configuration, and provide examples.

User asks about permissions: → Fetch /docs/en/permissions.md and /docs/en/permission-modes.md, explain permission levels and configuration.

User asks about CLI commands: → Fetch /docs/en/cli-reference.md, list relevant commands with flags and examples.

Important Notes

• The documentation is extensive (131+ pages). You don't need to fetch all pages — only the ones relevant to the question.
• When users ask about "how Claude Code works" or "what can Claude Code do", fetch the overview page.
• For troubleshooting, always check the official error reference and troubleshooting guides first.
• Claude Code documentation is actively maintained — prefer real-time fetch over cached knowledge.
• If a page doesn't exist or can't be fetched, fall back to your training knowledge but clearly note when the information may be outdated.