claude-code

name	claude-code
description	Expert reference for Claude Code itself — features, setup, troubleshooting, and enterprise use. Use this skill when answering questions about Claude Code installation, authentication, slash commands, agent skills, MCP server configuration, hooks, plugins, subagents/agent teams, sessions, IDE integration (VS Code, JetBrains, desktop, Chrome, Slack), CI/CD integration, enterprise deployment (SSO, sandboxing, monitoring), extended thinking, prompt caching, checkpointing, cost tracking, or debugging Claude Code behaviour. Distinct from the Agent SDK: this skill is about operating and understanding Claude Code as a tool, not about building applications on the Claude Agent SDK.
version	3.0.0
updated	"2026-06-01T00:00:00.000Z"
source	https://code.claude.com/docs/en/ (rebuilt from official llms-full.txt, 114 pages)

Claude Code Expert

Claude Code is Anthropic's agentic coding tool that lives in the terminal, IDE, desktop app, and browser. It understands a codebase, edits files, runs commands, and ships work faster — combining autonomous planning, execution, and validation with extensibility through skills, plugins, MCP servers, hooks, subagents, and agent teams.

This skill is the operating manual for Claude Code itself. It is not about the Claude Agent SDK (building apps on top of Claude Code as a library) — that has its own skill.

When to Use This Skill

Use when a user needs help with:

• Understanding Claude Code features, capabilities, and how the agent loop works
• Installation, setup, platforms, and authentication
• Slash commands, the CLI reference, interactive mode, and keybindings
• Creating or managing Agent Skills, plugins, and output styles
• Configuring MCP servers for external tool integration
• Setting up hooks (lifecycle automation)
• Subagents, agent teams, agent view, and parallel/dynamic workflows
• Sessions, checkpointing, permissions/permission modes, fast mode, plan/review modes
• Configuration: settings, env vars, model config, memory (CLAUDE.md), statusline
• IDE & app integrations (VS Code, JetBrains, desktop, Chrome, web, Slack, channels)
• Automation: GitHub Actions, GitLab CI/CD, headless mode, routines, scheduled tasks
• Deployment: Bedrock, Vertex AI, Foundry, sandboxing, devcontainers, network/LLM gateway
• Enterprise administration: admin setup, security, monitoring, costs, analytics, managed settings, compliance
• Troubleshooting and error resolution
• Recent changes (What's New / changelog)

Activation examples: "How do I use Claude Code?" · "What slash commands are available?" · "How to set up an MCP server?" · "Create a new skill for X" · "Fix Claude Code authentication issues" · "How do agent teams work?" · "Deploy Claude Code in an enterprise environment".

Core Concepts (at a glance)

• Subagents — specialized child agents with isolated context for parallel/focused work (references/subagents-and-teams.md).
• Agent teams / agent view — orchestrate and monitor many Claude Code sessions at once.
• Agent Skills — modular capabilities (SKILL.md + resources) Claude loads on demand; custom commands are now skills (references/skills.md).
• Slash commands — built-in + custom operations invoked with / (references/commands-and-cli.md).
• Hooks — shell commands that fire on lifecycle events (references/hooks.md).
• MCP servers — Model Context Protocol integrations for external tools (references/mcp.md).
• Plugins — packaged commands, skills, hooks, and MCP servers (references/plugins-and-output-styles.md).
• Memory — CLAUDE.md project/user instructions (references/configuration.md).

Reference Map (progressive disclosure)

Load a reference file only when the question calls for it. Each file is a focused slice of the official docs; grep within a file for specific terms rather than reading it whole.

Topic	Reference file	Load when the user asks about…
Getting started	references/getting-started.md	install, setup, quickstart, platforms, how it works, features overview, glossary
Workflows & best practices	references/common-workflows.md	common workflows, best practices, large codebases, worktrees, code review, context window, prompt library
Subagents & teams	references/subagents-and-teams.md	subagents, agents, agent teams, agent view, parallel work
Sessions & modes	references/sessions-and-modes.md	sessions, checkpointing, interactive/fullscreen, permissions & permission modes, plan/review (ultraplan/ultrareview), fast mode, voice, computer use, remote control, prompt caching
Agent Skills	references/skills.md	creating/using skills, SKILL.md format, invocation control, bundled skills
Plugins & output styles	references/plugins-and-output-styles.md	plugins, plugin marketplaces/dependencies/hints, plugins reference, output styles
Hooks	references/hooks.md	hook types, configuration, the hooks guide
MCP	references/mcp.md	MCP server setup, managed MCP
Slash commands & CLI	references/commands-and-cli.md	slash commands, CLI flags, keybindings, tools reference, statusline
Configuration	references/configuration.md	settings.json, env vars, terminal/model config, memory (CLAUDE.md), .claude/ directory
IDE & apps	references/ide-and-apps.md	VS Code, JetBrains, desktop, Chrome, web, Slack, channels, deep links
Automation & CI/CD	references/automation-cicd.md	GitHub Actions, GitLab CI/CD, headless, routines, scheduled tasks
Deployment & infra	references/deployment-and-infra.md	Bedrock, Vertex AI, Foundry, AWS, network config, LLM gateway, devcontainer, sandboxing
Enterprise & admin	references/enterprise-admin.md	admin setup, authentication, security, monitoring, costs, analytics, managed settings, compliance
Troubleshooting	references/troubleshooting.md	errors, troubleshooting, debugging your config
What's New & changelog	references/whats-new.md	recent features, weekly updates, release history

Full titled index of every source page: llms.txt.

Instructions for Claude

When answering a Claude Code question:

1. Identify the topic from the question and map it to a reference file using the table above.
2. Load only the relevant reference(s) — most questions need exactly one. Use grep/search within the file for the specific term (e.g., search references/hooks.md for PreToolUse) instead of reading the entire file.
3. Answer from the loaded reference, quoting exact flags, settings keys, file paths, and commands. Each page block in a reference keeps its canonical Source: URL — cite it when helpful.
4. Combine references only for genuinely cross-cutting questions (e.g., "secure enterprise setup with sandboxing and managed MCP" → enterprise-admin.md + deployment-and-infra.md + mcp.md).
5. Prefer current docs over memory. This skill was rebuilt from the official llms-full.txt; treat it as the source of truth and note when behaviour may be version-dependent.

Documentation links

• Docs: https://code.claude.com/docs/en/overview
• CLI reference: https://code.claude.com/docs/en/cli-reference
• What's new: https://code.claude.com/docs/en/whats-new
• Support: https://support.claude.com

Rebuilt with Skill Seekers from the official Claude Code documentation (llms-full.txt). Agent SDK pages are intentionally excluded — see the separate Agent SDK skill.