Use when asked to review all skills, generate a concise ordered summary of suggestions for improving each skill, and provide an overall assessment of the skills' effectiveness
Use when asked to review all skills, generate a concise ordered summary of suggestions for improving each skill, and provide an overall assessment of the skills' effectiveness
user-invocable
true
disable-model-invocation
true
1. Visibility:
Skills with high risk side effects (deploy, commit, send messages): add disable-model-invocation: true so Claude can't auto-fire.
Skills that are pure background knowledge users would never /run themselves: add user-invocable: false to hide from /menu.
2. Deterministic vs non-deterministic:
Find any step inside a skill where AI is interpreting something that's actually a fixed, repeatable operation.
Suggest replacing those steps with a script saved inside the skill folder.
Code = same result every time, no token cost. Keep AI for the steps that need judgment.
3. Composability:
Flag any skill that duplicates logic another skill already has.
Suggest extracting shared logic into a callable script or a smaller composable skill.
For each duplication found, propose a rewrite and explain what changed and why.