Claude Code SDK
Authoritative reference for Claude Code extensibility and configuration.
Route to Reference
| Working with | Reference |
|---|---|
| Skills: SKILL.md, frontmatter, activation, subagent execution | skills.md |
| Subagents: custom agents, tool restrictions, delegation | subagents.md |
| Plugins: plugin.json, directory structure, distribution | plugins.md |
| Hooks: lifecycle events, command/prompt/agent hooks | hooks.md |
| MCP: servers, tools, resources, authentication | mcp.md |
| Memory: CLAUDE.md, rules, imports, hierarchy | memory.md |
| Output Styles: personas, system prompt modification | output-styles.md |
| Settings: scopes, permissions, sandbox, env vars, tools | settings.md |
| Status line: custom display, JSON input, scripts | statusline.md |
| Usage patterns: context management, verification, scaling | best-practices.md |
Read the relevant reference before making changes.
Concepts
Plugin — Distributable package containing skills, hooks, MCP servers,
output styles. Has .claude-plugin/plugin.json manifest. Installed from
marketplace or local path. Skills namespaced as /plugin:skill.
Hook — Deterministic automation triggered at lifecycle events (tool use, session start, permission request). Three types: command (shell), prompt (LLM decision), agent (multi-turn verification). Configured in settings or plugin.
MCP Server — External tool/resource provider via Model Context Protocol. Connects Claude to databases, APIs, services. Configured per-project, per-user, or via plugin.
Output Style — Persona/behavior modifier via system prompt changes. Affects how Claude responds without changing capabilities. Built-in: Default, Explanatory, Learning.
CLAUDE.md — Project memory file providing persistent context about codebase, conventions, instructions. Hierarchy: managed → user → project. Loaded automatically at session start.
Subagent — Isolated context for delegated tasks. Built-in types:
Explore (read-only), Plan (architecture), general-purpose. Custom agents
in .claude/agents/.
Settings — Configuration hierarchy controlling permissions, model, hooks, behavior. Scopes: managed → user → project → local.
Quick Reference
Skill Locations
| Scope | Path |
|---|---|
| Personal | ~/.claude/skills/<name>/SKILL.md |
| Project | .claude/skills/<name>/SKILL.md |
| Plugin | <plugin>/skills/<name>/SKILL.md |
Plugin Layout
plugin-name/
├── .claude-plugin/
│ └── plugin.json # Manifest (required)
├── skills/
│ └── skill-name/
│ ├── SKILL.md
│ └── references/
├── hooks/
│ └── hooks.json
├── output-styles/
├── .mcp.json
└── README.md
Hook Events
| Event | When | Can Block |
|---|---|---|
PreToolUse | Before tool | Yes |
PostToolUse | After tool success | No |
Stop | Claude finishes | Yes |
SessionStart | Session begins | No |
Settings Scopes
| Scope | Location | Shared |
|---|---|---|
| Managed | System path | Org-wide |
| User | ~/.claude/settings.json | No |
| Project | .claude/settings.json | Git |
| Local | .claude/settings.local.json | No |
When to Use This Skill
- Creating plugins, skills, hooks, or output styles
- Debugging activation failures or hook issues
- Understanding Claude Code configuration hierarchy
- Setting up MCP servers or tool integration
- Writing or modifying CLAUDE.md files
- Any question about Claude Code internals