---

name: <kebab-case-skill-name> description: One paragraph — leads with "when to use" / trigger phrases. Stays ≤ 1536 chars. Example "Use when... User says... Auto-triggers on..." patterns work best. when_to_use: Additional trigger phrases the model uses to decide invocation. Concrete user prompts and phrases work better than abstract descriptions. argument-hint: "[arg-name]" allowed-tools: Read, Glob, Grep disable-model-invocation: false # true = user-only (no auto-trigger) user-invocable: true # false = hidden from / menu (Claude-only) model: inherit # opus | sonnet | haiku | inherit paths: # optional — auto-load only on matching files

• "src/**/*" context: fork # optional — runs in forked subagent (with agent: below) agent: <name> # required if context: fork mcpServers: [chrome-devtools] # optional — pre-attached MCP servers

---

<Skill Title>

One-sentence summary of what this skill does. Keep it crisp.

When to use

• Specific user phrase 1
• Specific user phrase 2
• Concrete scenario 1

When NOT to use

• Scenario where another skill is better
• Scenario where this is overkill

Process

1. Read inputs from <where>
2. Do <thing>
3. Validate via <how>
4. Output to <where>
5. Hand off to <next step>

Hard rules

• Rule 1 — why it matters
• Rule 2 — why it matters
• Rule 3 — why it matters

Auto-clarity escape hatch (per caveman pattern)

If the skill modifies tone or behavior, declare when it must DISABLE itself:

• Security warnings
• Irreversible action confirmations
• Multi-step sequences where fragment order matters
• Ambiguity needing clarification

Output

<path-or-artifact-this-produces>

After save: "<one-line summary of what was done>."

References

• Source / inspiration: <github URL>
• Related skills: .claude/skills/<other>/SKILL.md
• Anthropic skills spec: https://code.claude.com/docs/en/skills