template-complete.md 2.4 KB


name: {{skill_name}}

description: "[Domain] end-to-end capability: includes [capability 1], [capability 2], [capability 3]. Use when [decidable triggers]."

{{skill_name}} Skill

Production-grade skill for [domain]: extract rules, patterns, and reproducible examples from source material (avoid documentation dumps).

When to Use This Skill

Trigger when any of these applies:

  • You are designing/implementing/debugging [domain/tech]
  • You need to turn requirements into concrete commands/code/configs
  • You need common pitfalls, boundaries, and acceptance criteria

Not For / Boundaries

  • What this skill will not do (prevents misfires and over-promising)
  • Required inputs; ask 1-3 questions if missing

Quick Reference

Common Patterns

Pattern 1: one-line explanation

[command/snippet you can paste and run]

Pattern 2:

[command/snippet you can paste and run]

Rules & Constraints

  • MUST: non-negotiable rules (security boundaries, defaults, acceptance)
  • SHOULD: strong recommendations (best practices, performance habits)
  • NEVER: explicit prohibitions (dangerous ops, inventing facts)

Examples

Example 1

  • Input:
  • Steps:
  • Expected output / acceptance:

Example 2

Example 3

FAQ

  • Q: ...
    • A: ...

Troubleshooting

  • Symptom -> Likely causes -> Diagnosis -> Fix

References

  • references/index.md: navigation
  • references/getting_started.md: onboarding and vocabulary
  • references/api.md: API/CLI/config reference (if applicable)
  • references/examples.md: long examples and extra use cases
  • references/troubleshooting.md: edge cases and failure modes

Maintenance

  • Sources: docs/repos/specs (do not invent)
  • Last updated: YYYY-MM-DD
  • Known limits: what is explicitly out of scope

Quality Gate

Minimum checks before shipping (see meta-skill claude-skills for the full version):

  1. description is decidable ("what + when") and includes trigger keywords
  2. Has "When to Use This Skill" with decidable triggers
  3. Has "Not For / Boundaries" to reduce misfires
  4. Quick Reference is <= 20 patterns and each is directly usable
  5. Has >= 3 reproducible examples (input -> steps -> acceptance)
  6. Long content is in references/ with a navigable references/index.md
  7. Uncertain claims include a verification path (no bluffing)
  8. No documentation dumps in Quick Reference
  9. Reads like an operator's manual, not a knowledge dump