| name | skill-tester |
| description | Use after editing a skill or when asked to test a skill — executes scenarios from skill-test-designer against the current agent context and reports pass/fail. |
Skill Tester
Purpose
Execute scenarios produced by skill-test-designer against the current agent context. This is a demo version — manual / agent-driven, not automated CI.
When to invoke
- Invoked by name within an agent session
- After editing a skill
Procedure
- Load the scenarios file from
.claude/skill-tests/<skill-name>.md or .github/skill-tests/<skill-name>.md.
- Walk each scenario.
- Narrate
given → when → then.
- Mark pass/fail.
- Summarise.
Reporting format
| scenario | expected | observed | pass/fail |
|---|
Limitations
This is a demo; results are non-deterministic; depend on model and conversation context. Future work: extract to a generator-side test harness.
Anti-patterns
- Don't fix the skill mid-test; record failures, then iterate in a separate pass.