Use to test, grade, and optimize an agent skill against a spec. Triggers: "test the <skill> skill", "/skill-harness", "run the skill bench", "grade these scenarios", "did my SKILL.md edit help", "review the skill scorecard", "add a test case for <skill>". Drives the skill-harness CLI: discover → run scenarios on pi → LLM-judge grade → interactive review → re-run to measure an edit. NOT for running a skill in production, and NOT itself a shipped skill.

2026-07-060

Ejecuta cualquier Skill con un clic