name: claude-expert description: "Claude Code mastery skill. Use when configuring hooks, creating skills, orchestrating subagents, or checking for Claude Code updates. Triggers: 'create a hook', 'make a skill', 'spawn subagent', 'check for updates', 'update my config'."

Claude Code Expert

You are a Claude Code expert. Use this skill when working on Claude Code configuration.

Hook Events

Event	When	Use For
`SessionStart`	Session begins	Inject context, load state
`UserPromptSubmit`	Before user message processed	Validate, suggest skills
`PreToolUse`	Before tool executes	Guard, validate params
`PostToolUse`	After tool executes	Track changes, run checks
`PreCompact`	Before context compaction	Save state, create handoff
`Stop`	Agent completes	Cleanup, summarize
`SessionEnd`	Session ends	Final cleanup
`SubagentStop`	Subagent completes	Collect results
`Notification`	Background event	Alert user

Hook Structure

{
  "hooks": {
    "EventName": [
      {
        "matcher": "*",  // or "Bash", "Edit", etc.
        "hooks": [
          {
            "type": "command",
            "command": "path/to/script.sh",
            "timeout": 10000
          }
        ]
      }
    ]
  }
}

Skill Structure

skill-name/
├── SKILL.md          # Frontmatter + instructions
├── references/       # Lazy-loaded details
└── scripts/          # Executable helpers

Frontmatter:

---
name: skill-name
description: "When to use this skill. Triggers: 'keyword1', 'keyword2'."
---

Key: Description should say WHEN to use, not just what it does.

Subagent Patterns

When to spawn:

Task is self-contained and won't need back-and-forth
You need focused context (fresh context window)
Parallel work is possible

When NOT to spawn:

Simple queries (use Grep/Glob directly)
Need iterative refinement with user
Task requires current conversation context

Context hygiene:

Give subagent focused, minimal prompt
Request structured output (bullets, not prose)
Don't spawn subagents from subagents

Staying Current

Check for updates:

When you notice new features or deprecations, propose config updates.

Quick Reference

Create a hook: Add to settings.json, create script in hooks/ Create a skill: Make directory in skills/, write SKILL.md Create a command: Add .md file to commands/ with frontmatter Spawn subagent: Use Task tool with appropriate subagent_type

Name	claude-expert
Description	Claude Code mastery skill. Use when configuring hooks, creating skills, orchestrating subagents, or checking for Claude Code updates. Triggers: 'create a hook', 'make a skill', 'spawn subagent', 'check for updates', 'update my config'.