Builds a skill from scratch, edits one that already exists, and measures the result. The measurement half matters more than it sounds: a skill that never triggers is worse than no skill, because you think you have coverage you do not have. This runs evals against the description to check the trigger actually catches the phrasings it needs to.
Fires when creating a skill, editing one, optimizing a description, or benchmarking triggering accuracy.
This is the skill that builds the rest of the library on this page.