A collection of reusable extensions for Claude Code: agents, skills, hooks, and scripts.
- claude-sandbox - Containerized environment for running Claude Code in autonomous mode
- mcp-servers - MCP servers (protocol layer, portable across clients)
This repo covers Claude Code's extension points:
| Type | Description |
|---|---|
| Hooks | Event handlers that modify Claude's behavior |
| Skills | Reusable capabilities invoked via /skill-name (includes custom slash commands) |
| Agents | Specialized AI for specific tasks |
| Settings | Useful settings.json configurations (statusline, etc.) |
| Scripts | Standalone tools that complement Claude Code |
claude-code-extensions/
├── hooks/ # Event handlers
├── skills/ # Slash commands and auto-invoked capabilities
├── agents/ # Specialized AI for specific tasks
├── settings/ # Settings.json configurations
├── scripts/ # Standalone tools (session-resume, etc.)
├── README.md # This file
└── CLAUDE.md # Workflow and conventions
Using /sync (recommended): From this project directory, run /sync inside Claude Code. It syncs all extension types (agents, skills, hooks) to ~/.claude/, detects conflicts, and verifies hook enablement in settings.json.
Manual installation:
- Hooks: Copy scripts to
~/.claude/hooks/and add configuration to~/.claude/settings.json(seehooks/README.mdfor each hook's config) - Skills: Copy skill directories to
~/.claude/skills/ - Agents: Copy agent files to
~/.claude/agents/ - Settings: Copy scripts to
~/.claude/settings/and add configuration to~/.claude/settings.json(seesettings/README.md)
MIT