mcollina/skill-optimizer
Optimizes AI skills for activation, clarity, and cross-model reliability. Use when creating or editing skill packs, diagnosing weak skill uptake, reducing regressions, tuning instruction salience, improving examples, shrinking context cost, or setting benchmark and release gates for skills. Trigger terms: skill optimization, activation gap, benchmark skill, with/without skill delta, regression, context budget, prompt salience.
87
Quality
87%
Does it follow best practices?
Impact
Pending
No eval scenarios have been run
Securityby
Passed
No known issues
regression-triage.mdrules/
- name:
- regression-triage
- description:
- Diagnose and correct cases where skills reduce model performance
- metadata:
- {"tags":"regressions, triage, diagnostics, fixes"}
Regression triage
Detect
Flag any criterion/scenario where with skill < without skill.
Prioritize regressions on high-frequency tasks and strict-format outputs.
Classify
Common causes:
- ambiguous instruction collisions
- optional language around mandatory behavior
- over-broad rule that suppresses needed detail
- examples that imply the wrong default
Fix sequence
- isolate the exact criterion/regression scenario
- locate instruction lines likely causing confusion
- rewrite with explicit must/should boundaries
- add a small positive+negative example pair
- rerun the same scenario before broad reruns
Example fix pattern
Before:
- "Include references when relevant"
After:
- "If task requests a
Refs:footer, includeRefs: <url>exactly and do not omit it."
Exit criteria
- regression removed on affected models
- no new regressions in adjacent scenarios
- issue updated with root cause + diff summary