claude-code-features - SKILL.md Agent Skill

name: claude-code-features description: Comprehensive reference of Claude Code features. Use when asking about Claude Code capabilities, hooks, skills, agents, plugins, or configuration.

Claude Code Feature Reference

Comprehensive list of Claude Code features as of January 2026.

Configuration Files

File	Purpose
`CLAUDE.md`	Project memory, read at session start
`.claude/settings.json`	Hook configuration, permissions
`.claude/settings.local.json`	Personal overrides (gitignored)
`.mcp.json`	MCP server configuration
`.lsp.json`	LSP server configuration

Hooks (13 Types)

Hook	Trigger
`SessionStart`	Session begins
`SessionEnd`	Session ends
`UserPromptSubmit`	User submits prompt
`PreToolUse`	Before tool execution
`PostToolUse`	After tool succeeds
`PostToolUseFailure`	After tool fails
`PermissionRequest`	Permission requested
`Notification`	Notification sent
`Stop`	Agent stopping
`SubagentStart`	Subagent begins
`SubagentStop`	Subagent completes
`PreCompact`	Before context compaction
`Setup`	Repository init/maintenance

Hook Types

command - Execute shell command
prompt - Inject prompt (Stop/SubagentStop only)

Hook Response Format

{
  "continue": true,
  "stopReason": "string",
  "systemMessage": "string",
  "hookSpecificOutput": {
    "hookEventName": "PreToolUse",
    "permissionDecision": "allow|deny|ask"
  }
}

Skills

Required Frontmatter

name - Lowercase, hyphens, max 64 chars
description - When to use, max 1024 chars

Optional Frontmatter

allowed-tools - Tools allowed without permission
model - Specific model to use
context: fork - Isolated sub-agent context
agent - Agent type for fork context
hooks - Lifecycle hooks
user-invocable - Slash command visibility

Subagents

Required Frontmatter

name - Lowercase with hyphens
description - When to delegate

Optional Frontmatter

tools - Available tools
disallowedTools - Denied tools
model - sonnet, opus, haiku, inherit
permissionMode - default, acceptEdits, dontAsk, bypassPermissions, plan
skills - Skills to load
hooks - Lifecycle hooks

Built-in Agents

Explore - Fast read-only (Haiku)
Plan - Research for planning
general-purpose - Complex tasks
Bash - Terminal commands

Plugins

Structure

plugin/
├── .claude-plugin/plugin.json
├── commands/
├── agents/
├── skills/
├── hooks/hooks.json
├── .mcp.json
└── .lsp.json

Permissions

{
  "permissions": {
    "allow": ["Bash(git diff:*)"],
    "ask": ["Bash(git push:*)"],
    "deny": ["Read(./.env)"]
  }
}

MCP Servers

{
  "mcpServers": {
    "name": {
      "command": "npx",
      "args": ["-y", "@package/name"],
      "env": {}
    }
  }
}

Last verified: January 2026