| name | skill-name |
| description | Use when [specific trigger, symptom, user request, or workflow state] |
Skill Name
One or two sentences explaining the skill's purpose and core principle.
State the invariant this skill protects.
When to Use
Use this skill when:
- [Concrete user phrase, artifact state, symptom, or situation]
- [Another trigger future agents are likely to search for]
Do not use this skill for:
- [Nearby case that belongs to another skill or normal project instructions]
Process
- [First required action]
- [Second required action]
- [Verification or handoff action]
Write steps as commands to the agent. If a step depends on another file, name
that file and say exactly when to open it.
Stop and Ask
STOP and ask the user when:
- [A missing decision would change the implementation or output]
- [Continuing would require unsafe guessing, credentials, or destructive work]
Do not ask about facts that can be discovered by reading the repo.
Red Flags
| Thought | Reality |
|---|
| "[Tempting shortcut]" | [Why it fails and what to do instead] |
| "[Another predictable rationalization]" | [Correct behavior] |
Output
Return this shape:
[Expected final response, artifact, or report format]
If the skill produces a file, name the path pattern and the handoff skill or
next workflow step.
References
- supporting-file.md - when to open it
related-skill-name - when this skill should defer to it