What problem does it solve? Skill authors often change instructions based on preference rather than evidence, breaking routing or behavior without noticing. This Skill enforces an evidence-driven improvement cycle: baseline testing, targeted diagnosis, minimal patching, and re-validation with keep-or-revert decisions based on score changes. ## Core Features & Use Cases - Baseline Testing: Runs the companion skill-test workflow first and records failures before any edit is made. - Failure Diagnosis: Classifies each failure as routing ambiguity, missing workflow detail, unsafe mutation, broken reference, unsupported metadata, platform coupling, excessive context, or unverifiable output. - Approval-Gated Patching: Proposes the smallest patch with affected files, behavior changes, regressions, and rollback, then waits for user approval before editing. - Use Case: A skill triggers on the wrong prompts. Run this Skill to capture baseline test results, diagnose the routing ambiguity, apply a minimal frontmatter fix, re-run all direct, implicit, near-miss, and negative tests, and keep the change only if scores improve. ## Quick Start Ask Codex to improve a specific skill using the test-fix-retest loop and report the diff, validation evidence, and rollback command.
