By name: your-skill-name description: One or two sentences that say WHAT this skill does AND WHEN the agent should load it. Write in the third person, name concrete tasks, and include trigger keywords a user would actually say. Example: "Generate professional release notes from a git history. Use when cutting a release, drafting a changelog, or summarizing commits since the last tag."
Your Skill Name
Delete this template's comments as you fill it in. Keep the body under ~500 lines — push deep material into
references/, deterministic helpers intoscripts/, and templates/fonts/icons intoassets/. See../spec/README.mdfor the full standard.
When to use this
Describe the situations this skill is for, in one tight paragraph. Be specific — this
reinforces the description and helps the model trigger correctly.
Instructions
Give the agent clear, opinionated guidance: the workflow, the decisions, the gotchas. Prefer numbered steps for fragile or order-dependent procedures, prose for open ones. Assume the reader is a capable engineer — don't explain the basics.
- Step one.
- Step two.
- Step three.
References (optional)
Link to deeper docs the agent should read only when needed:
references/deep-dive.md— loaded on demand (link it once the file exists).
Scripts (optional)
Describe any executable helpers and whether the agent should run or read them:
scripts/do_thing.py— run to perform X deterministically.
Examples (optional)
Show one or two worked examples so the agent can pattern-match the desired output.