Template. Copy to
<target-repo>/.agents/skills/<repo>-<domain>/SKILL.mdand fill every bracketed placeholder from the live target repo. Delete this line and every other>guidance line before committing. See../../references/repo-local-skill-generation.mdfor the full process.
name: - description: | [One paragraph: what this skill verifies/runs/operates for , stated in terms of the repo's real shape (service/CLI/library/etc.), not generic process. End with explicit trigger phrasing: "Use when: '', ''. Trigger: /-."] argument-hint: "[||]"
-
[One or two sentences: what "" means in this repo specifically, and the one deterministic gate/command that is necessary-but-not-sufficient (name it, name what it can't prove).]
Surfaces
Only include rows for surfaces this repo actually has. Delete the rest. Every cell is a real path/command from the live repo, not a placeholder genre.
| Changed area | Surface | Verification path |
|---|---|---|
<path-glob> |
[HTTP service / CLI / library / worker / SDK] | [exact command(s)] |
Commands
Exact, copy-pasted invocations. Include the one-time setup a cold agent needs (env vars, local DB path, ports) and the fallback if the default port/resource is taken. If the repo's CI workflow differs from what AGENTS.md/README claims, name the mismatch here instead of picking one.
[real command 1]
[real command 2]
Gotchas
Repo-specific footguns a cold agent would otherwise rediscover the hard way — pulled from AGENTS.md/CLAUDE.md's own footgun list, docs/, or the generation author's own dry run. Not generic advice ("write tests").
- [gotcha 1]
- [gotcha 2]
Report
Return: verdict (PASS / FAIL / UNVERIFIED) · exact command(s) run · surface(s) exercised · artifact/output inspected · what was NOT covered.