template-skill

Copy this file to skills/<your-skill>/SKILL.md and replace each section.

<!-- DELETE everything between « » after filling in. These are authoring hints. -->

When To Use

<!-- «Scope anchor — required by contract. List 3-5 concrete situations.» -->

• «Situation where this skill adds value over general agent capability»
• «Second trigger scenario»
• «Third trigger scenario»

Boundaries

<!-- «Non-goals — required by contract. What this skill does NOT do.» -->

• «Task type this skill should not be used for»
• «Adjacent skill or workflow that handles the excluded case»
• Do not «specific anti-pattern to avoid»

Workflow

<!-- «Execution anchor — required by contract. Numbered steps or headed phases.» -->

1. «First step — what the agent does and with what input»
2. «Second step»
3. «Third step»
4. «Final step — what the agent produces or hands off»

Output

<!-- «Output anchor — required by contract. What the user gets.» -->

• «Primary deliverable (file, report, code, etc.)»
• «Secondary deliverable if any»

Verification

<!-- «Verification anchor — required by contract. How to confirm quality.» -->

• «Testable criterion for the primary output»
• «Second quality check»
• quick_validate.py passes on the skill directory

Resources

<!-- «Resource map — required by contract IF the skill bundles scripts/, references/, assets/, or commands/. Delete this section if no resources exist.» -->

• scripts/«name».sh — «what it does»
• references/«name».md — «what it contains»

---

Authoring Checklist

Use this checklist before shipping. All items are enforced by validate_skill_contract.py --strict:

• [ ] Frontmatter name is hyphen-case, matches directory name, max 64 chars
• [ ] Frontmatter description includes trigger language (Use when..., Triggers on...)
• [ ] Description is under 1024 chars, no angle brackets
• [ ] When To Use section present (scope anchor)
• [ ] Boundaries section present (boundaries anchor)
• [ ] Numbered steps or workflow heading present (execution anchor)
• [ ] Output section present (output anchor)
• [ ] Verification section present (verification anchor)
• [ ] Resources section present if skill has scripts/, references/, assets/, or commands/
• [ ] SKILL.md is under 500 lines (501-700 warns, >700 requires decomposition)
• [ ] All «placeholder» text replaced
• [ ] All <!-- comments --> removed
• [ ] Validated: python3 skills/skill-creator/scripts/quick_validate.py <skill-path>
• [ ] Strict pass: python3 skills/skill-evals/scripts/validate_skill_contract.py --skills <name> --strict

Sibling skills

Part of the skill-management toolchain.

• skill-creator — guided authoring path. Use that skill when you want full guidance; use this template when you just need a contract-passing shell.
• skill-evals — runs the strict validator referenced in the checklist above.
• audit-skill — security audit before publishing/installing.