[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-73424":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":9,"language":10,"languages":9,"totalLinesOfCode":9,"stars":11,"forks":12,"watchers":13,"openIssues":14,"contributorsCount":15,"subscribersCount":15,"size":15,"stars1d":16,"stars7d":17,"stars30d":18,"stars90d":15,"forks30d":15,"starsTrendScore":19,"compositeScore":20,"rankGlobal":9,"rankLanguage":9,"license":9,"archived":21,"fork":21,"defaultBranch":22,"hasWiki":23,"hasPages":21,"topics":24,"createdAt":9,"pushedAt":9,"updatedAt":25,"readmeContent":26,"aiSummary":27,"trendingCount":15,"starSnapshotCount":15,"syncStatus":28,"lastSyncTime":29,"discoverSource":30},73424,"claude-code-guide","Cranot\u002Fclaude-code-guide","Cranot","The Complete Claude Code CLI Guide - Live & Auto-Updated Every 2 Days",null,"Shell",2795,316,15,1,0,3,10,86,9,76.1,false,"main",true,[],"2026-06-12 04:01:09","# The Complete Claude Code CLI Guide\n\n[![Official Docs](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FOfficial_Docs-code.claude.com-blue)](https:\u002F\u002Fcode.claude.com\u002Fdocs\u002Fen\u002Foverview) [![GitHub](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FGitHub-anthropics%2Fclaude--code-black)](https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-code) [![NPM](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FNPM-@anthropic--ai%2Fclaude--code-red)](https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002F@anthropic-ai\u002Fclaude-code) [![Auto-Updated](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FAuto--Updated-Every%202%20Days-brightgreen)](#auto-update-pipeline)\n\n**Quick Links:** [Get Started](#what-is-claude-code) · [Commands](#quick-reference) · [MCP Setup](https:\u002F\u002Fcode.claude.com\u002Fdocs\u002Fen\u002Fmcp) · [Settings](https:\u002F\u002Fcode.claude.com\u002Fdocs\u002Fen\u002Fsettings) · [SDK](https:\u002F\u002Fcode.claude.com\u002Fdocs\u002Fen\u002Fsdk) · [Changelog](#changelog)\n\n> **🔄 Live Guide**: Auto-updated every 2 days from [official docs](https:\u002F\u002Fcode.claude.com\u002Fdocs\u002Fen\u002Foverview), [GitHub releases](https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-code\u002Freleases), and [Anthropic changelog](https:\u002F\u002Fwww.anthropic.com\u002Fchangelog). See [update-log.md](.\u002Fupdate-log.md).\n\n> **🤖 For AI Agents**: Optimized for both humans and AI. `[OFFICIAL]` = from code.claude.com. `[COMMUNITY]` = observed patterns. `[EXPERIMENTAL]` = unverified.\n\n---\n\n## What is Claude Code?\n\n**Claude Code is an agentic AI coding assistant that lives in your terminal.** It understands your codebase, edits files directly, runs commands, and helps you code faster through natural language conversation.\n\n**Key Capabilities:**\n- 💬 Natural language interface in your terminal\n- 📝 Direct file editing and command execution\n- 🔍 Full project context awareness\n- 🔗 External integrations via MCP (Model Context Protocol)\n- 🤖 Extensible via Skills, Hooks, and Plugins\n- 🛡️ Sandboxed execution for security\n\n**Installation:**\n```bash\n# Quick Install (macOS, Linux, WSL)\ncurl -fsSL https:\u002F\u002Fclaude.ai\u002Finstall.sh | bash\n\n# Windows PowerShell\nirm https:\u002F\u002Fclaude.ai\u002Finstall.ps1 | iex\n\n# Windows CMD\ncurl -fsSL https:\u002F\u002Fclaude.ai\u002Finstall.cmd -o install.cmd && install.cmd && del install.cmd\n\n# Alternative: Homebrew (macOS\u002FLinux)\nbrew install --cask claude-code\n\n# Alternative: WinGet (Windows)\nwinget install Anthropic.ClaudeCode\n\n# Alternative: NPM (⚠️ Deprecated - use native install instead)\nnpm install -g @anthropic-ai\u002Fclaude-code\n\nclaude --version  # Verify installation\n```\n\n**Official Documentation:** https:\u002F\u002Fcode.claude.com\u002Fdocs\u002Fen\u002Foverview\n\n---\n\n## Contents\n\n| Getting Started | Core Features | Practical Usage | Reference |\n|-----------------|---------------|-----------------|-----------|\n| [What is Claude Code?](#what-is-claude-code) | [Skills System](#skills-system) | [Development Workflows](#development-workflows) | [Security](#security-considerations) |\n| [Core Concepts](#core-concepts) | [Built-in Commands](#built-in-commands) | [Tool Synergies](#tool-synergies) | [SDK Integration](#sdk-integration) |\n| [Quick Start Guide](#quick-start-guide) | [Hooks System](#hooks-system) | [Examples Library](#examples-library) | [Troubleshooting](#troubleshooting) |\n| [Quick Reference](#quick-reference) | [MCP Integration](#mcp-integration) | [Best Practices](#best-practices) | [Changelog](#changelog) |\n| | [Sub-Agents](#sub-agents) | | [Auto-Update Pipeline](#auto-update-pipeline) |\n| | [Agent Teams](#agent-teams) | | |\n| | [Plugins](#plugins) | | |\n\n---\n\n## Quick Reference\n\n### Essential Commands [OFFICIAL]\n\n```bash\n# Starting Claude Code\nclaude                    # Start interactive session\nclaude -p \"task\"          # Print mode (non-interactive)\nclaude --continue         # Continue last session\nclaude --resume \u003Cid>      # Resume specific session\n\n# Session Management\n\u002Fhelp                     # Show available commands\n\u002Fexit                     # End session\n\u002Fcompact                  # Reduce context size\n\u002Fcompact [instructions]  # Compact conversation with optional focus instructions\n\n# Background Tasks\n\u002Fbashes                   # List background processes\n\u002Fkill \u003Cid>               # Stop background process\n\n# Discovery\n\u002Fcommands                 # List skills and commands\n\u002Fhooks                   # Show configured hooks\n\u002Fskills                  # List available Skills (NEW)\n\u002Fplugin                  # Manage plugins\n```\n\n**Source:** [CLI Reference](https:\u002F\u002Fcode.claude.com\u002Fdocs\u002Fen\u002Fcli-reference)\n\n### CLI Flags Reference [OFFICIAL]\n\n```bash\n# Output Control\nclaude -p, --print \"task\"          # Print mode: non-interactive, prints result and exits\nclaude --output-format json         # Output format: text, json, or stream-json\nclaude --input-format text          # Input format: text or stream-json\nclaude --verbose                    # Enable verbose logging (full turn-by-turn output)\n\n# Session Management\nclaude --continue                   # Continue from last session\nclaude --resume \u003Csession-id>        # Resume specific session by ID or name\nclaude --from-pr \u003Cpr>               # Resume session linked to GitHub PR number or URL [NEW]\nclaude --fork-session               # Create new session ID instead of reusing original\nclaude --session-id \u003Cuuid>          # Use specific session ID (must be valid UUID)\n\n# Remote Sessions (claude.ai subscribers)\nclaude --remote \"task\"              # Create web session on claude.ai\nclaude --teleport                   # Resume web session in local terminal\n\n# Debugging & Logging\nclaude --debug                      # Enable debug mode (with optional category filtering)\nclaude --debug \"api,mcp\"            # Debug specific categories\nclaude --debug \"!statsig,!file\"     # Exclude categories with !\n\n# Model & Agent Configuration\nclaude --model \u003Cname>               # Specify model (sonnet, opus, haiku, or full name)\nclaude --fallback-model \u003Cname>      # Fallback model when default overloaded (print mode)\nclaude --agent \u003Cname>               # Specify custom agent (overrides settings)\nclaude --agents '\u003Cjson>'            # Define custom subagents dynamically via JSON\n\n# System Prompt Customization\nclaude --system-prompt \"prompt\"     # Replace entire default system prompt\nclaude --system-prompt-file \u003Cpath>  # Replace with file contents (print mode only)\nclaude --append-system-prompt \"...\"  # Append to default system prompt\nclaude --append-system-prompt-file \u003Cpath>  # Append file contents (print mode only)\n\n# Tool & Permission Management\nclaude --tools \"Bash,Read,Edit\"     # Restrict built-in tools (use \"\" to disable all)\nclaude --allowedTools \"Bash(git:*)\" # Tools that execute without prompting\nclaude --disallowedTools \"Edit\"     # Tools removed from context\nclaude --permission-mode plan       # Begin in specified permission mode\nclaude --dangerously-skip-permissions  # Skip all permission prompts ⚠️\nclaude --allow-dangerously-skip-permissions  # Enable bypass option without activating [NEW]\nclaude --permission-prompt-tool \u003Cmcp-tool>  # MCP tool for permission prompts (non-interactive) [NEW]\n\n# Budget & Execution Limits (print mode)\nclaude --max-budget-usd 5.00        # Maximum dollar amount for API calls\nclaude --max-turns 3                # Limit number of agentic turns\nclaude --json-schema '\u003Cschema>'     # Get validated JSON output matching schema (print mode) [NEW]\n\n# Directory & Configuration\nclaude --add-dir ..\u002Fapps ..\u002Flib     # Add additional working directories\nclaude --plugin-dir .\u002Fmy-plugins    # Load plugins from directories\nclaude --settings .\u002Fsettings.json   # Path to settings JSON file\nclaude --setting-sources user,project  # Comma-separated list of setting sources [NEW]\nclaude --mcp-config .\u002Fmcp.json      # Load MCP servers from JSON file\nclaude --strict-mcp-config          # Only use MCP servers from --mcp-config\n\n# IDE & Browser Integration\nclaude --ide                        # Auto-connect to IDE on startup\nclaude --chrome                     # Enable Chrome browser integration\nclaude --no-chrome                  # Disable Chrome browser integration\n\n# Agent Teams [NEW]\nclaude --teammate-mode in-process   # Teammates display in main terminal\nclaude --teammate-mode tmux         # Each teammate in own pane (requires tmux\u002FiTerm2)\nclaude --teammate-mode auto         # Auto-detect (default)\n\n# Setup & Maintenance\nclaude --init                       # Run Setup hooks and start interactive mode\nclaude --init-only                  # Run Setup hooks and exit (no interactive session)\nclaude --maintenance                # Run Setup hooks with maintenance trigger and exit\n\n# Other Options\nclaude --disable-slash-commands     # Disable all skills and slash commands\nclaude --no-session-persistence     # Disable session persistence (print mode)\nclaude --betas interleaved-thinking # Beta headers for API requests\nclaude --include-partial-messages   # Include partial streaming events (with stream-json) [NEW]\n```\n\n**Common Flag Combinations:**\n\n```bash\n# One-off task with JSON output\nclaude --print \"analyze this code\" --output-format json\n\n# Debug MCP and API issues\nclaude --debug \"api,mcp\"\n\n# Resume session with specific model\nclaude --resume auth-refactor --model opus\n\n# Non-interactive with budget limit (CI\u002FCD)\nclaude -p --max-budget-usd 5.00 --output-format json \"run tests\"\n\n# Custom subagents for specialized work\nclaude --agents '{\"reviewer\":{\"description\":\"Code reviewer\",\"prompt\":\"Review for bugs\"}}'\n\n# Remote session for claude.ai subscribers\nclaude --remote \"fix the login bug\"\n```\n\n**Source:** [CLI Reference](https:\u002F\u002Fcode.claude.com\u002Fdocs\u002Fen\u002Fcli-reference)\n\n### Core Tools [OFFICIAL]\n\n| Tool | Purpose | Permission Required |\n|------|---------|---------------------|\n| **Read** | Read files, images, PDFs | No |\n| **Write** | Create new files | Yes |\n| **Edit** | Modify existing files | Yes |\n| **Bash** | Execute shell commands | Yes |\n| **Grep** | Search content with regex | No |\n| **Glob** | Find files by pattern | No |\n| **TodoWrite** | Task management | No |\n| **Task** | Launch sub-agents | No |\n| **WebFetch** | Fetch web content | Yes |\n| **WebSearch** | Search the web | Yes |\n| **NotebookEdit** | Edit Jupyter notebooks | Yes |\n| **NotebookRead** | Read Jupyter notebooks | No |\n\n**Source:** [Settings Reference](https:\u002F\u002Fcode.claude.com\u002Fdocs\u002Fen\u002Fsettings)\n\n---\n\n## Core Concepts\n\n### 1. How Claude Code Works [OFFICIAL]\n\nClaude Code operates through a **conversational interface** in your terminal:\n\n```bash\n# You describe what you want\n$ claude\n> \"Add user authentication to the API\"\n\n# Claude Code:\n1. Analyzes your codebase structure\n2. Plans the implementation\n3. Requests permission for file edits (first time)\n4. Writes code directly to your files\n5. Can run tests and verify changes\n6. Creates git commits if requested\n```\n\n**Key Principles:**\n- **Natural Language**: Just describe what you need - no special syntax\n- **Direct Action**: Edits files and runs commands with your permission\n- **Context Aware**: Understands your entire project structure\n- **Incremental Trust**: Asks permission as needed for new operations\n- **Scriptable**: Can be automated via SDK\n\n**Source:** [Overview](https:\u002F\u002Fcode.claude.com\u002Fdocs\u002Fen\u002Foverview)\n\n### 2. Permission Model [OFFICIAL]\n\nClaude Code uses an **incremental permission system** for safety:\n\n```bash\n# Permission Modes\n\"ask\"    # Prompt for each use (default for new operations)\n\"allow\"  # Permit without asking\n\"deny\"   # Block completely\n\n# Permission Priority [NEW v2.1.27]\n# Content-level rules override tool-level rules\n# Example: allow: [\"Bash\"], ask: [\"Bash(rm *)\"]\n#   -> Bash is generally allowed, but \"rm *\" commands require confirmation\n\n# Tools Requiring Permission\n- Bash (command execution)\n- Write\u002FEdit\u002FNotebookEdit (file modifications)\n- WebFetch\u002FWebSearch (network access)\n- Skill (skills and custom commands)\n\n# Tools Not Requiring Permission (Safe Operations)\n- Read\u002FNotebookRead (reading files)\n- Grep\u002FGlob (searching)\n- TodoWrite (task tracking)\n- Task (sub-agents)\n```\n\n**Configuring Permissions:**\n\nCreate `.claude\u002Fsettings.json` in your project or `~\u002F.claude\u002Fsettings.json` globally:\n\n```json\n{\n  \"permissions\": {\n    \"defaultMode\": \"ask\",\n    \"allow\": {\n      \"Bash\": [\"git status\", \"git diff\", \"git log\", \"npm test\", \"npm run*\"],\n      \"Read\": {},\n      \"Edit\": {}\n    },\n    \"deny\": {\n      \"Write\": [\"*.env\", \".env.*\", \".git\u002F*\"],\n      \"Edit\": [\"*.env\", \".env.*\"]\n    },\n    \"additionalDirectories\": [\n      \"\u002Fpath\u002Fto\u002Fother\u002Fproject\"\n    ]\n  }\n}\n```\n\n**Source:** [Settings](https:\u002F\u002Fcode.claude.com\u002Fdocs\u002Fen\u002Fsettings)\n\n### 3. Project Context - CLAUDE.md [COMMUNITY]\n\nA **CLAUDE.md** file in your project root provides persistent context across sessions:\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>Example CLAUDE.md file (click to expand)\u003C\u002Fstrong>\u003C\u002Fsummary>\n\n    # Project: My Application\n\n    ## Critical Context (Read First)\n    - Language: TypeScript + Node.js\n    - Framework: Express + React\n    - Database: PostgreSQL with Prisma ORM\n    - Testing: Jest + React Testing Library\n\n    ## Commands That Work\n    npm run dev          # Start dev server (port 3000)\n    npm test             # Run all tests\n    npm run lint         # ESLint check\n    npm run typecheck    # TypeScript validation\n    npm run db:migrate   # Run Prisma migrations\n\n    ## Important Patterns\n    - All API routes in \u002Fsrc\u002Froutes - RESTful structure\n    - Database queries use Prisma Client\n    - Auth uses JWT tokens (implementation in \u002Fsrc\u002Fauth)\n    - Frontend components in \u002Fsrc\u002Fcomponents\n    - API responses: {success: boolean, data: any, error?: string}\n\n    ## Gotchas & What NOT to Do\n    - DON'T modify \u002Fgenerated folder (auto-generated by Prisma)\n    - DON'T commit .env files (use .env.example instead)\n    - ALWAYS run npm run db:migrate after pulling schema changes\n    - DON'T use `any` type in TypeScript - use proper typing\n\n    ## File Structure\n    \u002Fsrc\n      \u002Froutes       # Express API routes\n      \u002Fservices     # Business logic\n      \u002Fmodels       # Type definitions\n      \u002Fmiddleware   # Express middleware\n      \u002Futils        # Shared utilities\n      \u002Fauth         # Authentication logic\n\n    ## Recent Learnings\n    - [2026-01-15] Payment webhook needs raw body parser for Stripe\n    - [2026-01-10] Redis pool: {maxRetriesPerRequest: 3}\n\n\u003C\u002Fdetails>\n\n**Why CLAUDE.md Helps:**\n- ✅ Provides context immediately at session start\n- ✅ Reduces need to re-explain project structure\n- ✅ Stores project-specific patterns and conventions\n- ✅ Documents what works (and what doesn't)\n- ✅ Shared with team via git\n- ✅ AI-optimized format for Claude to understand quickly\n\n**Note:** While CLAUDE.md is not an official feature, it's a widely-adopted community pattern. Claude Code will automatically read it if present at project root.\n\n### 4. Tools Reference [OFFICIAL]\n\n#### Read Tool\n**Purpose:** Read and analyze files\n\n```bash\n# Examples\nRead file_path=\"\u002Fsrc\u002Fapp.ts\"\nRead file_path=\"\u002Fdocs\u002Fscreenshot.png\"  # Can read images!\nRead file_path=\"\u002Fdocs\u002Fguide.pdf\"       # Can read PDFs!\nRead file_path=\"\u002Fdocs\u002Fguide.pdf\" pages=\"1-5\"  # Read specific PDF pages [NEW v2.1.30]\n```\n\n**Capabilities:**\n- Reads any text file (code, configs, logs, etc.)\n- Handles images (screenshots, diagrams, charts)\n- Processes PDFs - extracts text and visual content\n- Parses Jupyter notebooks (.ipynb files)\n- Returns content with line numbers (`cat -n` format)\n- Can read large files with offset\u002Flimit parameters\n\n**PDF Parameters** [NEW v2.1.30]:\n- `pages`: Optional page range (e.g., `\"1-5\"`, `\"1,3,5\"`) to read specific pages\n- Large PDFs (>10 pages) return a lightweight reference when @mentioned\n- PDF limits: Maximum 100 pages, 20MB file size\n\n**Special Features:**\n- **Images**: Claude can read screenshots of errors, UI designs, architecture diagrams\n- **PDFs**: Extract and analyze PDF content, useful for documentation and requirements\n- **Notebooks**: Full access to code cells, markdown, and outputs\n\n#### Write Tool\n**Purpose:** Create new files\n\n```bash\nWrite file_path=\"\u002Fsrc\u002FnewFile.ts\"\n      content=\"export const config = {...}\"\n```\n\n**Behavior:**\n- Creates new file with specified content\n- Will OVERWRITE if file already exists (use Edit for existing files)\n- Requires permission on first use per session\n- Creates parent directories if needed\n\n**Best Practice:** Use Edit tool for modifying existing files, Write tool only for new files.\n\n#### Edit Tool\n**Purpose:** Modify existing files with precise string replacement\n\n```bash\nEdit file_path=\"\u002Fsrc\u002Fapp.ts\"\n     old_string=\"const port = 3000\"\n     new_string=\"const port = process.env.PORT || 3000\"\n```\n\n**Important:**\n- Requires **exact string match** including whitespace and indentation\n- Fails if `old_string` is not unique in file (use larger context or `replace_all`)\n- Use `replace_all=true` to replace all occurrences (useful for renaming)\n- Must read file first before editing\n\n**Common Pattern:**\n```bash\n# 1. Read file to see exact content\nRead file_path=\"\u002Fsrc\u002Fapp.ts\"\n\n# 2. Edit with exact string match\nEdit file_path=\"\u002Fsrc\u002Fapp.ts\"\n     old_string=\"function login() {\n  return 'TODO';\n}\"\n     new_string=\"function login() {\n  return authenticateUser();\n}\"\n```\n\n#### Bash Tool\n**Purpose:** Execute shell commands\n\n```bash\nBash command=\"npm test\"\nBash command=\"git status\"\nBash command=\"find . -name '*.test.ts'\"\n```\n\n**Features:**\n- Can run any shell command\n- Supports background execution (`run_in_background=true`)\n- Configurable timeout (default 2 minutes, max 10 minutes)\n- Git operations are common (status, diff, log, commit, push)\n\n**Security:**\n- Requires permission\n- Can be restricted by pattern in settings\n- Sandboxing available on macOS\u002FLinux\n\n**Common Git Patterns:**\n```bash\n# Check status\nBash command=\"git status\"\n\n# View changes\nBash command=\"git diff\"\n\n# Create commit\nBash command='git add . && git commit -m \"feat: add authentication\"'\n\n# View history\nBash command=\"git log --oneline -10\"\n```\n\n#### Grep Tool\n**Purpose:** Search file contents with regex patterns\n\n```bash\n# Find functions\nGrep pattern=\"function.*auth\" path=\"src\u002F\" output_mode=\"content\"\n\n# Find TODOs with context\nGrep pattern=\"TODO\" output_mode=\"content\" -C=3\n\n# Count occurrences\nGrep pattern=\"import.*from\" output_mode=\"count\"\n\n# Case insensitive\nGrep pattern=\"error\" -i=true output_mode=\"files_with_matches\"\n```\n\n**Parameters:**\n- `pattern`: Regex pattern (ripgrep syntax)\n- `path`: Directory or file to search (default: current directory)\n- `output_mode`:\n  - `\"files_with_matches\"` (default) - Just file paths\n  - `\"content\"` - Show matching lines\n  - `\"count\"` - Show match counts per file\n- `-A`, `-B`, `-C`: Context lines (after, before, both)\n- `-i`: Case insensitive\n- `-n`: Show line numbers\n- `type`: Filter by file type (e.g., \"js\", \"py\", \"rust\")\n- `glob`: Filter by glob pattern (e.g., \"*.test.ts\")\n\n**Fast and Powerful:** Uses ripgrep under the hood, much faster than bash grep on large codebases.\n\n#### Glob Tool\n**Purpose:** Find files by pattern\n\n```bash\n# Find test files\nGlob pattern=\"**\u002F*.test.ts\"\n\n# Find specific extensions\nGlob pattern=\"src\u002F**\u002F*.{ts,tsx}\"\n\n# Find config files\nGlob pattern=\"**\u002Fconfig.{json,yaml,yml}\"\n```\n\n**Features:**\n- Fast pattern matching (works with any codebase size)\n- Returns files sorted by modification time (recent first)\n- Supports complex glob patterns (`**` for recursive, `{}` for alternatives)\n\n#### TodoWrite Tool\n**Purpose:** Manage task lists during work\n\n```bash\nTodoWrite todos=[\n  {\n    \"content\": \"Add authentication endpoint\",\n    \"status\": \"in_progress\",\n    \"activeForm\": \"Adding authentication endpoint\"\n  },\n  {\n    \"content\": \"Write integration tests\",\n    \"status\": \"pending\",\n    \"activeForm\": \"Writing integration tests\"\n  },\n  {\n    \"content\": \"Update API documentation\",\n    \"status\": \"pending\",\n    \"activeForm\": \"Updating API documentation\"\n  }\n]\n```\n\n**Task States:**\n- `\"pending\"` - Not started yet\n- `\"in_progress\"` - Currently working on (should be only ONE at a time)\n- `\"completed\"` - Finished successfully\n\n**Dependency Tracking** [NEW]: v2.1.16 introduced task dependency tracking, allowing tasks to define prerequisites that must complete before they start. This enables complex multi-step workflows with proper sequencing.\n\n**Best Practices:**\n- Use for multi-step tasks (3+ steps)\n- Keep ONE task `in_progress` at a time\n- Mark completed IMMEDIATELY after finishing\n- Use descriptive `content` (what to do) and `activeForm` (what you're doing)\n\n**When to Use:**\n- ✅ Complex multi-step features\n- ✅ User provides multiple tasks\n- ✅ Non-trivial work requiring planning\n- ❌ Single straightforward tasks\n- ❌ Trivial operations\n\n#### Task Tool (Sub-Agents)\n**Purpose:** Launch specialized AI agents for specific tasks\n\n```bash\n# Explore codebase\nTask subagent_type=\"Explore\"\n     prompt=\"Find all API endpoints and their authentication requirements\"\n\n# General purpose agent for complex tasks\nTask subagent_type=\"general-purpose\"\n     prompt=\"Research best practices for rate limiting APIs and implement a solution\"\n```\n\n**Available Sub-Agent Types:**\n- `\"general-purpose\"` - Complex multi-step tasks, research, implementation\n- `\"Explore\"` - Fast codebase exploration (Glob, Grep, Read, Bash)\n\n**When to Use:**\n- Research tasks requiring web search + analysis\n- Codebase exploration (finding patterns, understanding architecture)\n- Complex multi-step operations that can run independently\n- Background work while you continue other tasks\n\n#### WebFetch Tool\n**Purpose:** Fetch and analyze web page content\n\n```bash\nWebFetch url=\"https:\u002F\u002Fdocs.example.com\u002Fapi\"\n         prompt=\"Extract all endpoint documentation\"\n```\n\n**Features:**\n- Converts HTML to markdown for analysis\n- Can extract specific information with prompt\n- Useful for researching docs, articles, references\n\n#### WebSearch Tool\n**Purpose:** Search the web for current information\n\n```bash\nWebSearch query=\"React 19 new features 2024\"\n```\n\n**Use Cases:**\n- Research current best practices\n- Find up-to-date library documentation\n- Check for known issues or solutions\n- Verify latest framework features\n\n**Source:** [CLI Reference](https:\u002F\u002Fcode.claude.com\u002Fdocs\u002Fen\u002Fcli-reference), [Settings](https:\u002F\u002Fcode.claude.com\u002Fdocs\u002Fen\u002Fsettings)\n\n#### LSP Tool (Language Server Protocol) [OFFICIAL]\n**Purpose:** Get code intelligence features like go-to-definition, find references, and hover documentation.\n\n```bash\nLSP operation=\"goToDefinition\"\n    filePath=\"src\u002Futils\u002Fauth.ts\"\n    line=42\n    character=15\n```\n\n**Available Operations:**\n| Operation | Description |\n|-----------|-------------|\n| `goToDefinition` | Find where a symbol is defined |\n| `findReferences` | Find all references to a symbol |\n| `hover` | Get documentation and type info for a symbol |\n| `documentSymbol` | Get all symbols in a document (functions, classes, variables) |\n| `workspaceSymbol` | Search for symbols across the entire workspace |\n| `goToImplementation` | Find implementations of an interface or abstract method |\n| `prepareCallHierarchy` | Get call hierarchy item at a position |\n| `incomingCalls` | Find all functions\u002Fmethods that call the function at a position |\n| `outgoingCalls` | Find all functions\u002Fmethods called by the function at a position |\n\n**Parameters:**\n- `operation` (required): The LSP operation to perform\n- `filePath` (required): Absolute or relative path to the file\n- `line` (required): Line number (1-based, as shown in editors)\n- `character` (required): Character offset (1-based, as shown in editors)\n\n**Use Cases:**\n```bash\n# Find where a function is defined\n> \"Go to the definition of getUserById\"\n\n# Find all usages of a function\n> \"Find all references to the authenticate function\"\n\n# Get documentation for a symbol\n> \"What does the validateToken function do?\"\n\n# Explore code structure\n> \"List all symbols in the auth.ts file\"\n```\n\n**Note:** LSP servers must be configured for the file type. If no server is available for a language, an error will be returned.\n\n**Source:** [CLI Reference](https:\u002F\u002Fcode.claude.com\u002Fdocs\u002Fen\u002Fcli-reference)\n\n### 5. Context Management [OFFICIAL]\n\nClaude Code maintains conversation context with smart management:\n\n#### Context Commands\n\n```bash\n\u002Fcompact                   # Reduce context by removing old information\n\u002Fcompact \"keep auth work\"  # Compact with focus instructions (keeps specified context)\n```\n\n#### When to Use\n\n**Use \u002Fcompact when:**\n- Long sessions with many file reads\n- \"Context too large\" errors\n- You've completed a major task and want to start fresh\n\n**Use \u002Fcompact with instructions when:**\n- Context is getting large but you want to preserve recent work\n- Switching between related tasks\n- You want intelligent cleanup without losing important context\n- Example: `\u002Fcompact \"keep the authentication implementation context\"`\n\n#### What Gets Preserved vs Cleared\n\n**Preserved:**\n- CLAUDE.md content (your project context)\n- Recent interactions and decisions\n- Current task information and todos\n- Recent file reads still relevant\n\n**Cleared:**\n- Old file reads no longer needed\n- Completed operations\n- Stale search results\n- Old context no longer relevant\n\n#### Automatic Context Management\n\nClaude Code may automatically compact when:\n- Token limit is approaching\n- Many old file reads are present\n- Session has been very long\n\n**Source:** [Settings](https:\u002F\u002Fcode.claude.com\u002Fdocs\u002Fen\u002Fsettings)\n\n### 6. Workspace Management [OFFICIAL]\n\n#### Adding Directories with \u002Fadd-dir\n\nClaude Code can work with multiple directories simultaneously:\n\n```bash\n# Add another directory to current session\n\u002Fadd-dir \u002Fpath\u002Fto\u002Fother\u002Fproject\n\n# Work across multiple projects\n> \"Update the User type in backend and propagate to frontend\"\n# Claude can now access both directories\n```\n\n**Use Cases:**\n- Monorepo development (frontend + backend + shared libs)\n- Cross-project refactoring\n- Dependency updates across multiple projects\n- Coordinating changes between related repositories\n\n**Configuration:**\n\nYou can also pre-configure additional directories in `.claude\u002Fsettings.json`:\n\n```json\n{\n  \"permissions\": {\n    \"additionalDirectories\": [\n      \"\u002Fpath\u002Fto\u002Ffrontend\",\n      \"\u002Fpath\u002Fto\u002Fbackend\",\n      \"\u002Fpath\u002Fto\u002Fshared-libs\"\n    ]\n  }\n}\n```\n\n#### Status Line Configuration with \u002Fstatusline\n\nCustomize what information appears in your status line:\n\n```bash\n# Configure status line\n\u002Fstatusline\n\n# Options typically include:\n# - Current model\n# - Token usage\n# - Session duration\n# - Active tools\n# - Background processes\n```\n\n**Benefits:**\n- Monitor token usage in real-time\n- Track session duration\n- See active background processes\n- Understand which tools are being used\n\n**Source:** [CLI Reference](https:\u002F\u002Fcode.claude.com\u002Fdocs\u002Fen\u002Fcli-reference)\n\n---\n\n## Quick Start Guide\n\n### Your First Session\n\n```bash\n# 1. Navigate to your project\ncd \u002Fpath\u002Fto\u002Fyour\u002Fproject\n\n# 2. Start Claude Code\nclaude\n\n# 3. Ask Claude to understand your project\n> \"Read the codebase and explain the project structure\"\n\n# Claude will:\n- Look for README, package.json, or similar entry points\n- Read relevant files (asks permission first time)\n- Analyze the code structure\n- Provide a summary\n\n# 4. Request an analysis\n> \"Review the authentication system for security issues\"\n\n# Claude will:\n- Find authentication-related files\n- Analyze the implementation\n- Identify potential vulnerabilities\n- Suggest improvements\n\n# 5. Make changes\n> \"Add rate limiting to the login endpoint\"\n\n# Claude will:\n- Plan the implementation\n- Show you what changes will be made\n- Request permission to edit files\n- Implement the changes\n- Can run tests to verify\n\n# 6. Create a commit\n> \"Create a git commit for these changes\"\n\n# Claude will:\n- Run git status to see changes\n- Review git diff\n- Create a descriptive commit message\n- Commit the changes\n```\n\n### Setting Up Your Project for Claude Code\n\n#### 1. Create CLAUDE.md [COMMUNITY]\n\nThis provides context that persists across all sessions:\n\n```bash\n# Ask Claude to help create it\n> \"Create a CLAUDE.md file documenting this project's structure, commands, and conventions\"\n\n# Or create manually with:\n- Languages and frameworks used\n- Important commands (dev, test, build, lint)\n- Project structure overview\n- Coding conventions\n- Known gotchas or issues\n```\n\n#### 2. Configure Permissions (Optional) [OFFICIAL]\n\nCreate `.claude\u002Fsettings.json` in your project:\n\n```json\n{\n  \"permissions\": {\n    \"defaultMode\": \"ask\",\n    \"allow\": {\n      \"Bash\": [\n        \"npm test\",\n        \"npm run*\",\n        \"git status\",\n        \"git diff\",\n        \"git log*\"\n      ],\n      \"Read\": {},\n      \"Grep\": {},\n      \"Glob\": {}\n    },\n    \"deny\": {\n      \"Write\": [\"*.env\", \".env.*\"],\n      \"Edit\": [\"*.env\", \".env.*\", \".git\u002F*\"]\n    }\n  }\n}\n```\n\nThis configuration:\n- Allows common safe commands without asking\n- Blocks editing sensitive files\n- Still asks permission for file modifications\n\n#### 3. Test the Setup\n\n```bash\n> \"Run the tests\"\n# Should execute without permission prompt (if configured)\n\n> \"What commands are available?\"\n# Claude will read package.json and list scripts\n\n> \"What's in CLAUDE.md?\"\n# Claude will read and summarize your project context\n```\n\n**Source:** [Quickstart](https:\u002F\u002Fcode.claude.com\u002Fdocs\u002Fen\u002Fquickstart), [Settings](https:\u002F\u002Fcode.claude.com\u002Fdocs\u002Fen\u002Fsettings)\n\n---\n\n## Advanced Features\n\n### Thinking Mode [OFFICIAL]\n\nClaude Code supports extended thinking for complex reasoning tasks. Opus 4.5 has thinking mode enabled by default.\n\n**Activation Methods:**\n\n```bash\n# Toggle with keyboard shortcut\nAlt+T (or Option+T on macOS)  # Toggle thinking on\u002Foff\n\n# Or use natural language\n> \"think about this problem\"\n> \"think harder about the architecture\"\n> \"ultrathink about this security issue\"\n\n# Tab key (sticky toggle)\nPress Tab to toggle thinking mode on\u002Foff for subsequent prompts\n```\n\n**Thinking Levels:**\n| Trigger | Thinking Budget | Use Case |\n|---------|----------------|----------|\n| `think` | Standard | General reasoning, code analysis |\n| `think harder` | Extended | Complex problems, multiple approaches |\n| `ultrathink` | Maximum | Critical decisions, deep architecture analysis |\n\n**Best Practices:**\n- Use `think harder` for debugging complex issues\n- Use `ultrathink` for architectural decisions or security reviews\n- Thinking content is visible in `Ctrl+O` transcript mode\n- Thinking mode is sticky - stays on until toggled off\n\n**Source:** [Thinking Mode](https:\u002F\u002Fcode.claude.com\u002Fdocs\u002Fen\u002Fthinking-mode)\n\n### Fast Mode [NEW] [OFFICIAL]\n\nFast mode is a high-speed configuration for Claude Opus 4.6, making responses **2.5x faster** at a higher cost per token. Available since v2.1.36.\n\n**Toggle Fast Mode:**\n```bash\n# Toggle with built-in command\n\u002Ffast          # Toggle on\u002Foff\n\n# Or set in settings\n\"fastMode\": true   # In user settings file\n```\n\n**Visual Indicators:**\n- `↯` icon appears next to prompt when fast mode is active\n- Icon turns gray during rate limit cooldown\n\n**Pricing (per MTok):**\n| Mode | Input (\u003C200K) | Output | Input (>200K) | Output |\n|------|--------------|--------|---------------|--------|\n| Standard Opus 4.6 | $15 | $75 | $15 | $75 |\n| Fast Mode | $30 | $150 | $60 | $225 |\n\n**Note:** Fast mode is available at 50% discount until February 16, 2026.\n\n**Requirements:**\n- Claude subscription plan (Pro\u002FMax\u002FTeam\u002FEnterprise) or Claude Console API\n- Extra usage enabled (`\u002Fextra-usage`)\n- Not available on third-party providers (Bedrock, Vertex, Azure Foundry)\n- For Teams\u002FEnterprise: Admin must enable in organization settings\n\n**When to Use:**\n- ✅ Rapid iteration on code changes\n- ✅ Live debugging sessions\n- ✅ Time-sensitive work\n- ❌ Long autonomous tasks (cost matters more)\n- ❌ Batch processing or CI\u002FCD pipelines\n\n**Fast Mode vs Effort Level:**\n| Setting | Effect |\n|---------|--------|\n| **Fast mode** | Same quality, lower latency, higher cost |\n| **Lower effort level** | Faster responses, potentially lower quality |\n\nYou can combine both for maximum speed on straightforward tasks.\n\n**Rate Limits:**\n- Separate rate limits from standard Opus 4.6\n- Automatically falls back to standard mode during cooldown\n- Re-enables when cooldown expires\n\n**Source:** [Fast Mode](https:\u002F\u002Fcode.claude.com\u002Fdocs\u002Fen\u002Ffast-mode)\n\n### Plan Mode [OFFICIAL]\n\nPlan Mode provides structured planning with model selection for complex tasks.\n\n```bash\n# Enter plan mode\n\u002Fplan\n\n# Or Claude may suggest plan mode for complex tasks\n> \"Implement a complete authentication system\"\n# Claude: \"This is a complex task. Would you like me to create a plan first?\"\n```\n\n**Plan Mode Features:**\n- **Opus planning, Sonnet execution** - Uses stronger model for planning, faster model for implementation\n- **SonnetPlan Mode** - Sonnet planning, Haiku execution (cost-effective)\n- **Shift+Tab** - Auto-accept edits in plan mode\n- **Plan persistence** - Plans persist across `\u002Fclear`\n\n**Plan Mode Workflow:**\n1. Claude analyzes the task and creates a structured plan\n2. You review and approve or modify the plan\n3. Claude executes the plan step by step\n4. Progress is tracked with TodoWrite\n\n**Source:** [Plan Mode](https:\u002F\u002Fcode.claude.com\u002Fdocs\u002Fen\u002Fplan-mode)\n\n### Background Tasks & Agents [OFFICIAL]\n\nRun commands and agents in the background while continuing to work.\n\n**Keyboard Shortcut:**\n```bash\nCtrl+B  # Background current command or agent (unified shortcut)\n```\n\n**Background Commands:**\n```bash\n# Start command in background\n> \"Run the dev server in background\"\n> \"Start tests in watch mode in background\"\n\n# Or prefix with &\n> \"& npm run dev\"\n\n# View background tasks\n\u002Ftasks\n\u002Fbashes\n\n# Kill a background task\n\u002Fkill \u003Ctask-id>\n```\n\n**Background Agents:**\n```bash\n# Launch agent in background\n> \"Have an Explore agent analyze the codebase architecture in background\"\n\n# Agents run asynchronously and notify you when complete\n# You receive wake-up messages when background agents finish\n```\n\n**Features:**\n- Real-time output streaming to status line\n- Wake-up notifications when tasks complete\n- Multiple concurrent background processes\n- Output persisted to files for large outputs\n\n**Source:** [Background Tasks](https:\u002F\u002Fcode.claude.com\u002Fdocs\u002Fen\u002Fbackground-tasks)\n\n### Auto-Memory [NEW]\n\nClaude Code now automatically records and recalls memories as it works (v2.1.32+).\n\n**How It Works:**\n- Claude automatically remembers important context, decisions, and patterns\n- Memories persist across sessions and inform future work\n- No manual intervention required\n\n**Memory Scopes for Agents:**\n```markdown\n---\nname: my-agent\nmemory: project  # Options: user, project, local\n---\n```\n\n| Scope | Storage | Shared |\n|-------|---------|--------|\n| `user` | `~\u002F.claude\u002F` | All your projects |\n| `project` | `.claude\u002F` | Team via git |\n| `local` | `.claude\u002F*.local.*` | No (gitignored) |\n\n**Disable Auto-Memory:**\n```bash\nexport CLAUDE_CODE_DISABLE_AUTO_MEMORY=1\n```\n\n### Keyboard Shortcuts [OFFICIAL]\n\n**Navigation & Editing:**\n| Shortcut | Action |\n|----------|--------|\n| `Ctrl+R` | Search command history |\n| `Ctrl+O` | View transcript (shows thinking blocks) |\n| `Ctrl+G` | Edit prompt in system text editor |\n| `Ctrl+Y` | Readline-style paste (yank) |\n| `Alt+Y` | Yank-pop (cycle through kill ring) |\n| `Ctrl+B` | Background current command\u002Fagent |\n| `Ctrl+Z` | Suspend\u002FUndo |\n\n**Model & Mode Switching:**\n| Shortcut | Action |\n|----------|--------|\n| `Alt+P` (Win\u002FLinux) \u002F `Option+P` (macOS) | Switch models while typing |\n| `Alt+T` (Win\u002FLinux) \u002F `Option+T` (macOS) | Toggle thinking mode |\n| `Tab` | Toggle thinking (sticky) \u002F Accept suggestions |\n| `Shift+Tab` | Auto-accept edits (plan mode) \u002F Switch modes (Windows) |\n\n**Input & Submission:**\n| Shortcut | Action |\n|----------|--------|\n| `Enter` | Submit prompt \u002F Accept suggestion immediately |\n| `Shift+Enter` | New line (works in iTerm2, WezTerm, Ghostty, Kitty) |\n| `Tab` | Edit\u002Faccept prompt suggestion |\n| `Ctrl+T` | Toggle syntax highlighting in `\u002Ftheme` |\n\n**Image & File Handling:**\n| Shortcut | Action |\n|----------|--------|\n| `Cmd+V` (macOS) \u002F `Alt+V` (Windows) | Paste image from clipboard |\n| `Cmd+N` \u002F `Ctrl+N` | New conversation (VSCode) |\n\n**Vim Bindings (if enabled):**\n| Shortcut | Action |\n|----------|--------|\n| `;` and `,` | Repeat last motion |\n| `y` | Yank operator |\n| `p` \u002F `P` | Paste |\n| `Alt+B` \u002F `Alt+F` | Word navigation |\n\n**Login & Authentication:**\n| Shortcut | Action |\n|----------|--------|\n| `c` | Copy OAuth URL during login |\n\n**Bash Mode Autocomplete** [NEW v2.1.14]:\n| Shortcut | Action |\n|----------|--------|\n| `!` + `Tab` | History-based autocomplete - complete partial commands from history |\n\n### Prompt Suggestions [OFFICIAL]\n\nClaude Code suggests prompts based on context (enabled by default).\n\n```bash\n# Claude suggests contextual prompts\n> _  # Cursor blinking\n# Suggestion appears: \"Review the changes we made\"\n\n# Tab to edit the suggestion\nTab → Edit the suggestion text\n\n# Enter to submit immediately\nEnter → Submit the suggestion as-is\n```\n\n**Configuration:**\n```bash\n# Toggle in \u002Fconfig\n\u002Fconfig\n# Search for \"prompt suggestions\"\n# Toggle enable\u002Fdisable\n```\n\n### Environment Variables [OFFICIAL]\n\n**Core Configuration:**\n| Variable | Description |\n|----------|-------------|\n| `ANTHROPIC_API_KEY` | Your API key |\n| `CLAUDE_CODE_SHELL` | Override shell detection |\n| `CLAUDE_CODE_TMPDIR` | Custom temp directory |\n| `CLAUDE_CODE_DISABLE_BACKGROUND_TASKS` | Disable background task system |\n| `CLAUDE_CODE_ENABLE_TASKS` | Set to `false` to use legacy task system [NEW v2.1.19] |\n\n**Display & UI:**\n| Variable | Description |\n|----------|-------------|\n| `CLAUDE_CODE_HIDE_ACCOUNT_INFO` | Hide account info in UI |\n\n**Bash & Commands:**\n| Variable | Description |\n|----------|-------------|\n| `BASH_DEFAULT_TIMEOUT_MS` | Default bash command timeout |\n| `BASH_MAX_TIMEOUT_MS` | Maximum allowed timeout |\n| `CLAUDE_BASH_NO_LOGIN` | Don't use login shell |\n| `CLAUDE_BASH_MAINTAIN_PROJECT_WORKING_DIR` | Keep working directory |\n| `CLAUDE_CODE_SHELL_PREFIX` | Prefix for shell commands |\n\n**Model Configuration:**\n| Variable | Description |\n|----------|-------------|\n| `ANTHROPIC_DEFAULT_SONNET_MODEL` | Override default Sonnet model |\n| `ANTHROPIC_DEFAULT_OPUS_MODEL` | Override default Opus model |\n| `ANTHROPIC_DEFAULT_HAIKU_MODEL` | Override default Haiku model |\n| `ANTHROPIC_LOG` | Enable debug logging |\n\n**MCP Configuration:**\n| Variable | Description |\n|----------|-------------|\n| `MCP_TIMEOUT` | MCP connection timeout |\n| `MCP_TOOL_TIMEOUT` | Individual tool timeout |\n\n**File & Context:**\n| Variable | Description |\n|----------|-------------|\n| `CLAUDE_CODE_FILE_READ_MAX_OUTPUT_TOKENS` | Max tokens for file reads |\n| `CLAUDE_CODE_ADDITIONAL_DIRECTORIES_CLAUDE_MD` | Set to `1` to load CLAUDE.md from `--add-dir` directories [NEW] |\n| `CLAUDE_PROJECT_DIR` | Override project directory |\n| `CLAUDE_PLUGIN_ROOT` | Plugin root substitution |\n| `CLAUDE_CONFIG_DIR` | Custom config directory |\n| `XDG_CONFIG_HOME` | XDG config base path |\n\n**Network & Proxy:**\n| Variable | Description |\n|----------|-------------|\n| `NODE_EXTRA_CA_CERTS` | Custom CA certificates |\n| `NO_PROXY` | Proxy bypass list |\n| `CLAUDE_CODE_PROXY_RESOLVES_HOSTS` | Proxy DNS resolution |\n\n**Auto-Update & Plugins:**\n| Variable | Description |\n|----------|-------------|\n| `DISABLE_AUTOUPDATER` | Disable auto-updates |\n| `FORCE_AUTOUPDATE_PLUGINS` | Force plugin updates |\n| `CLAUDE_CODE_EXIT_AFTER_STOP_DELAY` | Exit delay after stop |\n\n**Monitoring & Telemetry:**\n| Variable | Description |\n|----------|-------------|\n| `CLAUDE_CODE_ENABLE_TELEMETRY` | Enable OpenTelemetry collection (`1`) |\n| `OTEL_METRICS_EXPORTER` | OTel metrics exporter (e.g., `otlp`) |\n| `DISABLE_TELEMETRY` | Opt out of Statsig telemetry (`1`) |\n| `DISABLE_ERROR_REPORTING` | Opt out of Sentry error reporting (`1`) |\n| `DISABLE_COST_WARNINGS` | Disable cost warning messages (`1`) |\n\n**Advanced:**\n| Variable | Description |\n|----------|-------------|\n| `CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS` | Disable anthropic-beta headers (workaround for gateway users) |\n| `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS` | Enable agent teams feature (`1`) [NEW] |\n| `CLAUDE_CODE_DISABLE_AUTO_MEMORY` | Disable automatic memory recording (`1`) [NEW] |\n| `CLAUDE_CODE_DISABLE_BACKGROUND_TASKS` | Disable background task system (`1`) |\n| `DISABLE_INTERLEAVED_THINKING` | Disable interleaved thinking |\n| `USE_BUILTIN_RIPGREP` | Use built-in ripgrep |\n| `CLOUD_ML_REGION` | Cloud ML region for Vertex |\n| `AWS_BEARER_TOKEN_BEDROCK` | AWS bearer token |\n| `MAX_THINKING_TOKENS` | Extended thinking budget (default: 31,999) |\n| `MAX_MCP_OUTPUT_TOKENS` | Max MCP tool response tokens (default: 25,000) |\n| `CLAUDE_CODE_MAX_OUTPUT_TOKENS` | Max output tokens (default: 32,000, max: 64,000) |\n| `CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC` | Disable autoupdate, bug reporting, telemetry |\n\n### New Settings [OFFICIAL]\n\nRecent settings additions (configure in `\u002Fconfig` or `settings.json`):\n\n```json\n{\n  \u002F\u002F Response language\n  \"language\": \"en\",  \u002F\u002F Claude's response language\n\n  \u002F\u002F Git integration\n  \"attribution\": true,  \u002F\u002F Add model name to commit bylines\n  \"respectGitignore\": true,  \u002F\u002F Respect .gitignore in searches\n\n  \u002F\u002F UI preferences\n  \"showTurnDuration\": true,  \u002F\u002F Show turn duration messages\n  \"fileSuggestion\": \"custom-cmd\",  \u002F\u002F Custom @ file search command\n  \"spinnerVerbs\": [\"analyzing\", \"thinking\", \"processing\"],  \u002F\u002F Custom spinner verbs\n  \"prefersReducedMotion\": false,  \u002F\u002F Reduce UI animations for accessibility [NEW v2.1.30]\n\n  \u002F\u002F Session behavior\n  \"companyAnnouncements\": true,  \u002F\u002F Show startup announcements\n\n  \u002F\u002F Plan mode\n  \"plansDirectory\": \".claude\u002Fplans\"  \u002F\u002F Custom directory for plan files\n}\n```\n\n**Skills Variable Substitution:** [NEW]\n```markdown\n# In skill files, use ${CLAUDE_SESSION_ID} for session-specific operations\nSession ID: ${CLAUDE_SESSION_ID}\n```\n\n**Project Rules:**\n```bash\n# New: .claude\u002Frules\u002F directory for project-specific rules\n.claude\u002Frules\u002F\n├── coding-style.md      # Coding conventions\n├── testing.md           # Testing requirements\n└── security.md          # Security guidelines\n```\n\n**Wildcard Permissions:**\n```json\n{\n  \"permissions\": {\n    \"allow\": {\n      \"Bash\": [\"npm *\", \"git *\"],  \u002F\u002F Wildcard patterns\n      \"mcp__myserver__*\": {}       \u002F\u002F MCP tool wildcards\n    }\n  }\n}\n```\n\n---\n\n## Skills System\n\n**Skills are unified capabilities that extend Claude Code — both auto-activated by Claude and manually invoked via `\u002Fskill-name`.**\n\n> **Note:** Custom slash commands (`.claude\u002Fcommands\u002F` files) have been merged into skills as of v2.1.3. Your existing command files keep working unchanged. Skills are recommended for new work because they support additional features like supporting files, invocation control, and subagent execution. See [Migration: Commands to Skills](#migration-commands-to-skills).\n\nClaude Code skills follow the [Agent Skills](https:\u002F\u002Fagentskills.io) open standard, which works across multiple AI tools. Claude Code extends the standard with additional features like invocation control, subagent execution, and dynamic context injection.\n\n### What Are Skills? [OFFICIAL]\n\nSkills are instructions packaged as `SKILL.md` files that extend what Claude Code can do. Claude loads them when relevant to your request, or you invoke them directly:\n\n```\n# Claude auto-activates a skill based on your request\nYou: \"Review this code for security issues\"\nClaude: [Loads security-reviewer skill automatically]\n\n# Or you invoke a skill directly\nYou: \u002Fsecurity-reviewer src\u002Fauth.ts\nClaude: [Loads and executes the security-reviewer skill]\n```\n\n**Two types of skill content:**\n\n- **Reference content** — Knowledge Claude applies to your current work (conventions, patterns, style guides). Runs inline alongside your conversation context.\n- **Task content** — Step-by-step instructions for a specific action (deploy, commit, code generation). Often invoked manually with `\u002Fskill-name`.\n\n### Where Skills Live [OFFICIAL]\n\nWhere you store a skill determines who can use it:\n\n| Location | Path | Applies To |\n|----------|------|------------|\n| **Enterprise** | [Managed settings](https:\u002F\u002Fcode.claude.com\u002Fdocs\u002Fen\u002Fpermissions#managed-settings) | All users in organization |\n| **Personal** | `~\u002F.claude\u002Fskills\u002F\u003Cskill-name>\u002FSKILL.md` | All your projects |\n| **Project** | `.claude\u002Fskills\u002F\u003Cskill-name>\u002FSKILL.md` | This project only |\n| **Plugin** | `\u003Cplugin>\u002Fskills\u002F\u003Cskill-name>\u002FSKILL.md` | Where plugin is enabled |\n\nWhen skills share the same name, higher-priority locations win: **Enterprise > Personal > Project**. Plugin skills use a `plugin-name:skill-name` namespace, so they cannot conflict.\n\n**Legacy compatibility:** Files in `.claude\u002Fcommands\u002F` still work and support the same frontmatter. If a skill and a command share the same name, the skill takes precedence.\n\n**Automatic nested directory discovery:** When you work with files in subdirectories, Claude Code discovers skills from nested `.claude\u002Fskills\u002F` directories. For example, editing a file in `packages\u002Ffrontend\u002F` also loads skills from `packages\u002Ffrontend\u002F.claude\u002Fskills\u002F`. This supports monorepo setups where packages have their own skills.\n\n**Live change detection:** Skills from directories added via `--add-dir` are loaded automatically and picked up by live change detection — edit them during a session without restarting.\n\n### Skill Directory Structure [OFFICIAL]\n\nEach skill is a directory with `SKILL.md` as the entrypoint:\n\n```\nmy-skill\u002F\n├── SKILL.md           # Main instructions (required)\n├── template.md        # Template for Claude to fill in (optional)\n├── examples\u002F\n│   └── sample.md      # Example output (optional)\n└── scripts\u002F\n    └── validate.sh    # Script Claude can execute (optional)\n```\n\nReference supporting files from your `SKILL.md` so Claude knows what each file contains:\n\n```markdown\n## Additional resources\n- For complete API details, see [reference.md](reference.md)\n- For usage examples, see [examples.md](examples.md)\n```\n\n> **Tip:** Keep `SKILL.md` under 500 lines. Move detailed reference material to separate files.\n\n### Creating a Skill [OFFICIAL]\n\n**Step 1:** Create the skill directory:\n```bash\n# Personal skill (available in all projects)\nmkdir -p ~\u002F.claude\u002Fskills\u002Fexplain-code\n\n# Project skill (shared with team via git)\nmkdir -p .claude\u002Fskills\u002Fexplain-code\n```\n\n**Step 2:** Write `SKILL.md` with frontmatter and instructions:\n```yaml\n---\nname: explain-code\ndescription: Explains code with visual diagrams and analogies. Use when explaining how code works, teaching about a codebase, or when the user asks \"how does this work?\"\n---\n\nWhen explaining code, always include:\n\n1. **Start with an analogy**: Compare the code to something from everyday life\n2. **Draw a diagram**: Use ASCII art to show the flow, structure, or relationships\n3. **Walk through the code**: Explain step-by-step what happens\n4. **Highlight a gotcha**: What's a common mistake or misconception?\n\nKeep explanations conversational. For complex concepts, use multiple analogies.\n```\n\n**Step 3:** Test the skill:\n```bash\n# Let Claude invoke it automatically\n> \"How does this code work?\"\n\n# Or invoke it directly\n> \u002Fexplain-code src\u002Fauth\u002Flogin.ts\n```\n\n### Frontmatter Reference [OFFICIAL]\n\nConfigure skill behavior with YAML frontmatter between `---` markers at the top of `SKILL.md`. All fields are optional; only `description` is recommended.\n\n| Field | Required | Description |\n|-------|----------|-------------|\n| `name` | No | Display name. If omitted, uses directory name. Lowercase letters, numbers, hyphens (max 64 chars). |\n| `description` | Recommended | What the skill does and when to use it. Claude uses this to decide when to load it. |\n| `argument-hint` | No | Hint shown during autocomplete (e.g., `[issue-number]` or `[filename] [format]`). |\n| `disable-model-invocation` | No | `true` → only user can invoke via `\u002Fname`. Default: `false`. |\n| `user-invocable` | No | `false` → hidden from `\u002F` menu, only Claude can invoke. Default: `true`. |\n| `allowed-tools` | No | Tools Claude can use without asking permission when skill is active. |\n| `model` | No | Model to use when skill is active. |\n| `context` | No | Set to `fork` to run in a forked subagent context. |\n| `agent` | No | Which subagent type to use when `context: fork` is set. |\n| `hooks` | No | Hooks scoped to this skill's lifecycle. See [Hooks](https:\u002F\u002Fcode.claude.com\u002Fdocs\u002Fen\u002Fhooks#hooks-in-skills-and-agents). |\n\n### Controlling Invocation [OFFICIAL]\n\nBy default, both you and Claude can invoke any skill. Two frontmatter fields restrict this:\n\n- **`disable-model-invocation: true`** — Only you can invoke. Use for workflows with side effects (e.g., `\u002Fdeploy`, `\u002Fcommit`).\n- **`user-invocable: false`** — Only Claude can invoke. Use for background knowledge that isn't actionable as a command.\n\n```yaml\n# User-only skill (Claude won't auto-trigger)\n---\nname: deploy\ndescription: Deploy the application to production\ndisable-model-invocation: true\n---\n\n# Model-only skill (hidden from \u002F menu)\n---\nname: legacy-system-context\ndescription: Background knowledge about the legacy system\nuser-invocable: false\n---\n```\n\n**Invocation and context-loading behavior:**\n\n| Frontmatter | You Can Invoke | Claude Can Invoke | When Loaded into Context |\n|-------------|----------------|-------------------|--------------------------|\n| (default) | Yes | Yes | Description always in context; full skill loads when invoked |\n| `disable-model-invocation: true` | Yes | No | Description not in context; full skill loads when you invoke |\n| `user-invocable: false` | No | Yes | Description always in context; full skill loads when invoked |\n\n**Restricting Claude's access via `\u002Fpermissions`:**\n\n```bash\n# Allow only specific skills\nSkill(commit)\nSkill(review-pr *)\n\n# Deny specific skills\nSkill(deploy *)\n\n# Disable all skills\nSkill    # Add to deny rules\n```\n\nPermission syntax: `Skill(name)` for exact match, `Skill(name *)` for prefix match with any arguments.\n\n### Passing Arguments [OFFICIAL]\n\nSkills accept arguments via placeholder substitutions:\n\n| Variable | Description |\n|----------|-------------|\n| `$ARGUMENTS` | All arguments passed when invoking the skill |\n| `$ARGUMENTS[N]` | Specific argument by 0-based index (e.g., `$ARGUMENTS[0]`) |\n| `$N` | Shorthand for `$ARGUMENTS[N]` (e.g., `$0`, `$1`) |\n| `${CLAUDE_SESSION_ID}` | Current session ID (useful for logging) |\n\n**Example:**\n```yaml\n---\nname: fix-issue\ndescription: Fix a GitHub issue\ndisable-model-invocation: true\n---\n\nFix GitHub issue $ARGUMENTS following our coding standards.\n\n1. Read the issue description\n2. Implement the fix\n3. Write tests\n4. Create a commit\n```\n\n```bash\n\u002Ffix-issue 123\n# Claude receives: \"Fix GitHub issue 123 following our coding standards...\"\n```\n\n**Indexed arguments:**\n```yaml\n---\nname: compare-files\ndescription: Compare two files\n---\n\n# Compare: $ARGUMENTS[0] vs $ARGUMENTS[1]\n# Shorthand: $0 vs $1\n\nCompare $0 and $1 for differences.\n```\n\n```bash\n\u002Fcompare-files \"src\u002Fv1\u002Fapi.ts\" \"src\u002Fv2\u002Fapi.ts\"\n# $0 = \"src\u002Fv1\u002Fapi.ts\", $1 = \"src\u002Fv2\u002Fapi.ts\"\n```\n\nIf `$ARGUMENTS` is not present in the skill content, arguments are appended as `ARGUMENTS: \u003Cvalue>`.\n\n### Advanced Patterns [OFFICIAL]\n\n#### Dynamic Context Injection\n\nThe `` !`command` `` syntax runs shell commands before the skill content is sent to Claude. The output replaces the placeholder:\n\n```yaml\n---\nname: pr-summary\ndescription: Summarize changes in a pull request\ncontext: fork\nagent: Explore\nallowed-tools: Bash(gh *)\n---\n\n## Pull request context\n- PR diff: !`gh pr diff`\n- PR comments: !`gh pr view --comments`\n- Changed files: !`gh pr diff --name-only`\n\n## Your task\nSummarize this pull request...\n```\n\nEach `` !`command` `` executes immediately (before Claude sees anything). Claude only sees the final result with actual data.\n\n#### Running in a Subagent\n\nAdd `context: fork` to run a skill in isolation. The skill content becomes the prompt that drives the subagent (no access to conversation history):\n\n```yaml\n---\nname: deep-research\ndescription: Research a topic thoroughly\ncontext: fork\nagent: Explore\n---\n\nResearch $ARGUMENTS thoroughly:\n\n1. Find relevant files using Glob and Grep\n2. Read and analyze the code\n3. Summarize findings with specific file references\n```\n\nThe `agent` field specifies which subagent to use. Options: built-in agents (`Explore`, `Plan`, `general-purpose`) or custom subagents from `.claude\u002Fagents\u002F`. Default: `general-purpose`.\n\n> **Warning:** `context: fork` only makes sense for skills with explicit instructions. Guidelines without a task will return without meaningful output.\n\n#### Extended Thinking\n\nTo enable extended thinking in a skill, include the word `ultrathink` anywhere in your skill content:\n\n```yaml\n---\nname: architecture-review\ndescription: Deep architectural analysis\n---\n\nUse ultrathink to analyze the architecture deeply.\n\nReview the overall structure, identify patterns, and suggest improvements.\n```\n\n### Practical Examples\n\n**Example: Code Review Skill**\n\n`.claude\u002Fskills\u002Fcode-reviewer\u002FSKILL.md`:\n```yaml\n---\nname: code-reviewer\ndescription: Reviews code for security vulnerabilities, bugs, performance issues, and style problems. Use when user asks to review, audit, or check code quality.\nallowed-tools: [Read, Grep, Glob]\n---\n\n# Code Review Skill\n\n## When to Activate\nUse this skill when the user asks to:\n- Review code for issues\n- Audit security or find vulnerabilities\n- Check code quality or best practices\n\n## Review Process\n\n### 1. Scope Detection\n- Use Glob to identify files to review\n- Prioritize recently modified files\n- Focus on user-specified areas if mentioned\n\n### 2. Analysis Layers\n- **Security**: SQL injection, XSS, auth issues, exposed secrets\n- **Bugs**: Logic errors, null checks, error handling\n- **Performance**: N+1 queries, unnecessary loops, memory leaks\n- **Style**: Naming conventions, code organization, readability\n\n### 3. Reporting\nProvide structured feedback organized by severity:\n- **Critical\u002FHigh**: Security issues\n- **Medium**: Performance issues\n- **Low**: Style and best practices\n\nEach issue: file path, description, and fix suggestion.\n```\n\n**Example: Test Generator Skill**\n\n`.claude\u002Fskills\u002Ftest-generator\u002FSKILL.md`:\n```yaml\n---\nname: test-generator\ndescription: Generates comprehensive unit and integration tests. Use when user asks to write tests, add test coverage, or create test cases.\nallowed-tools: [Read, Write, Grep, Glob, Bash]\n---\n\n# Test Generator Skill\n\n## When to Activate\nUse this skill when user requests:\n- \"Write tests for...\"\n- \"Add test coverage\"\n- \"Generate test cases\"\n\n## Test Generation Process\n\n### 1. Analyze Target Code\n- Read the file\u002Ffunction to test\n- Identify inputs, outputs, side effects\n- Check existing test patterns\n\n### 2. Generate Comprehensive Tests\nCover all scenarios:\n- Happy path (expected usage)\n- Error cases (invalid inputs)\n- Edge cases (empty, null, boundary values)\n- Side effects (database, API calls)\n\n### 3. Follow Project Patterns\n- Check CLAUDE.md for testing conventions\n- Match existing test file structure\n- Use project's test framework\n```\n\n**Example: Security Review Skill**\n\n`.claude\u002Fskills\u002Fsecurity-review\u002FSKILL.md`:\n```yaml\n---\nname: security-review\ndescription: Comprehensive security audit of codebase. Use when asked to review security, audit vulnerabilities, or check for exploits.\nallowed-tools: [Read, Grep, Glob]\ndisable-model-invocation: true\n---\n\n# Security Review: $ARGUMENTS\n\nPerform a thorough security audit focusing on: $ARGUMENTS\n\n## Review Checklist\n\n### 1. Authentication & Authorization\n- Check for weak password policies\n- Verify JWT token validation\n- Review session management\n- Check for broken access control\n\n### 2. Input Validation\n- SQL injection vulnerabilities\n- XSS (Cross-Site Scripting) risks\n- Command injection possibilities\n- Path traversal vulnerabilities\n\n### 3. Data Protection\n- Sensitive data exposure\n- Encryption at rest and in transit\n- API keys and secrets in code\n- Database credential security\n\n### 4. Dependencies\n- Known vulnerabilities in packages\n- Outdated dependencies\n- License compliance issues\n\n### 5. Configuration\n- Security headers (CSP, HSTS, etc.)\n- CORS configuration\n- Error messages leaking information\n- Debug mode in production\n\n**Output Format** - Provide a detailed report with sections:\n- Critical Issues (Fix Immediately)\n- High Priority\n- Medium Priority\n- Low Priority \u002F Recommendations\n- Security Strengths\n- Action Plan (prioritized list of fixes)\n```\n\n**Usage:**\n```bash\n\u002Fsecurity-review \"authentication and API endpoints\"\n```\n\n**Example: API Documentation Generator Skill**\n\n`.claude\u002Fskills\u002Fapi-docs\u002FSKILL.md`:\n```yaml\n---\nname: api-docs\ndescription: Generate comprehensive API documentation from code. Use when asked to document APIs, create API docs, or generate OpenAPI specs.\nallowed-tools: [Read, Write, Grep, Glob]\ndisable-model-invocation: true\n---\n\n# Generate API Documentation\n\nAnalyze the codebase and create comprehensive API documentation for: $ARGUMENTS\n\n## Process\n\n### 1. Discovery\n- Find all API routes\u002Fendpoints\n- Identify request\u002Fresponse types\n- Note authentication requirements\n- Document query parameters\n\n### 2. Documentation\nFor each endpoint, document:\n- Method and path\n- Description\n- Authentication requirements\n- Request body\u002Fparameters\n- Response codes and bodies\n- Example requests\n\n### 3. Output\n- Create `\u002Fdocs\u002FAPI.md` with full documentation\n- Create `\u002Fopenapi.yaml` with OpenAPI spec if applicable\n```\n\n**Usage:**\n```bash\n\u002Fapi-docs \"all endpoints\"\n\u002Fapi-docs \"authentication routes\"\n```\n\n### File References with @ Syntax [OFFICIAL]\n\nReference files with `@` prefix for quick file inclusion:\n\n```bash\n# Reference single file\n\u002Freview-code @src\u002Fauth.ts\n\n# Reference multiple files\n\u002Freview-code @src\u002Fauth.ts @src\u002Fapi.ts @tests\u002Fauth.test.ts\n\n# Works in regular prompts too\n> \"Review @src\u002Fservices\u002Fpayment.ts for security issues\"\n\n# Reference files with skill arguments\n\u002Fanalyze-file @src\u002Fcomponents\u002FUserProfile.tsx\n```\n\n**How @ References Work:**\n- `@filename` automatically expands to include file content\n- Works with both absolute and relative paths\n- Can reference multiple files in one command\n- Files are read and included in context automatically\n- Reduces need to explicitly say \"read file X first\"\n\n**Use Cases:**\n```bash\n# Code review with context\n> \"Compare @src\u002Fapi\u002Fv1.ts and @src\u002Fapi\u002Fv2.ts and list differences\"\n\n# Refactoring across files\n> \"Make @src\u002Fmodels\u002FUser.ts consistent with @src\u002Ftypes\u002Fuser.d.ts\"\n\n# Bug investigation\n> \"This error occurs in @src\u002Fservices\u002Fauth.ts, check @logs\u002Ferror.log for clues\"\n\n# Test generation\n> \"Generate tests for @src\u002Futils\u002Fvalidator.ts\"\n```\n\n**Best Practices:**\n- Use @ references when you know exact file paths\n- Combine with skills for reusable workflows\n- Great for focused analysis of specific files\n- Reduces token usage vs. reading entire directories\n\n### MCP Integration [OFFICIAL]\n\nMCP servers can expose prompts that become invocable skills automatically:\n\n```json\n{\n  \"prompts\": [\n    {\n      \"name\": \"search-docs\",\n      \"description\": \"Search internal documentation\",\n      \"arguments\": [{\"name\": \"query\", \"description\": \"Search query\"}]\n    }\n  ]\n}\n```\n\nThis becomes available as `\u002Fsearch-docs` in Claude Code.\n\n```bash\n# Add MCP server\nclaude mcp add github -- gh-mcp\n\n# MCP prompts become skills:\n\u002Fgithub-pr-review      # Review current PR\n\u002Fgithub-issues         # List open issues\n\u002Fgithub-create-pr      # Create PR from current branch\n```\n\n### Skill Best Practices [OFFICIAL]\n\n#### 1. Write Clear, Specific Descriptions\n\nThe `description` field is critical — it helps Claude decide when to activate:\n\n**Good:**\n```yaml\ndescription: \"Generates API documentation from code comments. Use when user asks to document APIs, create API docs, update endpoint documentation, or generate OpenAPI specs.\"\n```\n\n**Bad:**\n```yaml\ndescription: \"Documentation generator\"  # Too vague\n```\n\n#### 2. Use Natural Trigger Words\n\nInclude terms users would naturally say:\n\n```yaml\n# For security review skill\ndescription: \"Reviews code for security. Use when asked to: review security, audit code, find vulnerabilities, check for exploits, analyze risks.\"\n\n# For performance optimization skill\ndescription: \"Optimizes code performance. Use when asked to: improve performance, optimize speed, reduce memory usage, make faster, profile code.\"\n```\n\n#### 3. Restrict Tools Appropriately\n\n```yaml\n# Analysis only (can't modify code)\nallowed-tools: [Read, Grep, Glob]\n\n# Can create\u002Fmodify code\nallowed-tools: [Read, Write, Edit, Bash]\n\n# Research and implementation\nallowed-tools: [Read, Write, Edit, WebFetch, WebSearch]\n```\n\n#### 4. Keep Skills Focused\n\n**Good (focused):**\n- `sql-optimizer` — Optimizes SQL queries only\n- `api-docs-generator` — Generates API documentation\n- `security-scanner` — Finds security issues\n\n**Bad (too broad):**\n- `database-everything` — Too vague\n- `code-helper` — What kind of help?\n\n#### 5. Provide Clear Instructions\n\nStructure your SKILL.md:\n1. **When to Activate** — Clear triggers\n2. **Process** — Step-by-step what to do\n3. **Output Format** — How to present results\n4. **Examples** — Show expected behavior\n\n#### 6. Mind the Context Budget\n\nSkill descriptions are loaded into context so Claude knows what's available. If you have many skills, they may exceed the character budget (2% of context window, fallback 16,000 characters). Run `\u002Fcontext` to check for warnings about excluded skills.\n\nOverride the limit with the `SLASH_COMMAND_TOOL_CHAR_BUDGET` environment variable.\n\n### Troubleshooting Skills [OFFICIAL]\n\n**Skill not triggering:**\n1. Check the description includes keywords users would naturally say\n2. Verify the skill appears when you ask \"What skills are available?\"\n3. Try rephrasing your request to match the description\n4. Invoke directly with `\u002Fskill-name` to confirm it works\n\n**Skill triggers too often:**\n1. Make the description more specific\n2. Add `disable-model-invocation: true` for manual-only invocation\n\n**Claude doesn't see all skills:**\n- Too many skill descriptions may exceed the character budget\n- Run `\u002Fcontext` to check for a warning about excluded skills\n- Set `SLASH_COMMAND_TOOL_CHAR_BUDGET` to a higher value\n\n### Migration: Commands to Skills\n\nCustom slash commands (`.claude\u002Fcommands\u002F` files) have been merged into the skills system. **Your existing command files keep working unchanged.** Skills are recommended for new work because they support:\n\n- **Supporting files** — Bundle templates, scripts, and reference docs alongside your skill\n- **Invocation control** — Choose whether you, Claude, or both can invoke\n- **Subagent execution** — Run skills in isolated forked contexts\n- **Nested discovery** — Automatic loading from subdirectories (monorepo support)\n\n**Migration path:**\n```bash\n# Old structure (still works)\n.claude\u002Fcommands\u002Freview.md\n\n# New structure (recommended)\n.claude\u002Fskills\u002Freview\u002FSKILL.md\n```\n\nBoth create `\u002Freview` and work the same way. If both exist, the skill takes precedence.\n\n**Source:** [Agent Skills](https:\u002F\u002Fcode.claude.com\u002Fdocs\u002Fen\u002Fskills)\n\n---\n\n## Built-in Commands\n\n**Built-in commands are native CLI commands for managing your Claude Code session.** They are hardcoded into Claude Code and are NOT skills — you cannot customize or override them.\n\n> **Note:** For custom workflow commands, use [Skills](#skills-system) instead. Built-in commands like `\u002Fhelp` and `\u002Fcompact` are not available through the Skill tool.\n\n### Command Reference [OFFICIAL]\n\n```bash\n# Session Management\n\u002Fhelp              # Show all available commands\n\u002Fexit              # End current session\n\u002Fclear             # Clear conversation history\n\u002Fcompact [instr]   # Compact context (optionally specify what to focus on)\n\u002Frewind            # Undo code changes in conversation\n\n# Session & History\n\u002Frename \u003Cname>     # Give current session a name\n\u002Fresume [name|id]  # Resume a previous session by name or ID\n\u002Fexport            # Export conversation to file\n\u002Fcopy              # Copy last response to clipboard [NEW]\n\n# Usage & Stats\n\u002Fusage             # View plan limits and usage (NEW)\n\u002Fstats             # Usage stats, engagement metrics (supports 7\u002F30\u002Fall-time) (NEW)\n\u002Fextra-usage       # Enable extra usage for Max plan subscribers [NEW]\n\u002Ffast              # Toggle fast mode (uses faster model, available after \u002Fextra-usage) [NEW]\n\n# Background Process Management\n\u002Fbashes            # List all background processes\n\u002Ftasks             # List all background tasks (agents, shells, etc.)\n\u002Fkill \u003Cid>         # Stop a background process\n\n# Discovery & Debugging\n\u002Fbug               # Report bugs (sends conversation to Anthropic)\n\u002Fcommands          # List all skills and commands\n\u002Fdebug             # Troubleshoot session issues [NEW v2.1.30]\n\u002Fhooks             # Show configured hooks\n\u002Fskills            # List available Skills\n\u002Fplugin            # Plugin management interface\n\u002Fcontext           # View current context usage as a colored grid\n\u002Fcost              # Show token usage statistics\n\u002Fdoctor            # Run diagnostics (shows Updates section with auto-update channel)\n\n# Configuration\n\u002Fconfig            # General settings (type to search and filter)\n\u002Fpermissions       # Manage tool permissions (with search)\n\u002Fprivacy-settings  # View and update privacy settings\n\u002Fstatus            # Show session status (Status tab)\n\u002Fstatusline        # Configure status line display\n\u002Fmodel             # Switch between models\n\u002Foutput-style      # Set output style directly or from selection menu\n\u002Ftheme             # Theme picker (Ctrl+T toggles syntax highlighting)\n\u002Fterminal-setup    # Configure terminal (Kitty, Alacritty, Zed, Warp)\n\u002Fvim               # Enter vim mode for alternating insert\u002Fcommand modes\n\u002Fsandbox           # Enable sandboxed bash with filesystem\u002Fnetwork isolation\n\n# Workspace Management\n\u002Fadd-dir \u003Cpath>    # Add additional directory to workspace\n\u002Fagents            # Manage custom AI subagents\n\u002Finit              # Initialize project with CLAUDE.md guide\n\u002Fmemory            # Edit CLAUDE.md memory files\n\u002Finstall-github-app # Set up Claude GitHub Actions","Claude Code 是一个基于终端的AI编码助手，旨在通过自然语言对话帮助用户更快地编写代码。它支持直接文件编辑、命令执行，并具备全项目上下文感知能力，同时可以通过MCP协议实现外部集成。此外，Claude Code还提供了技能、钩子和插件扩展机制，以及安全沙箱执行环境来保障安全性。该项目特别适合需要提高开发效率、希望通过自然语言与代码库进行交互的开发者使用。其自动更新机制确保了文档每两天同步一次官方最新内容，保证了信息的时效性和准确性。",2,"2026-06-11 03:45:29","high_star"]