What problem does it solve?
Accumulated memory feedback about a Claude Code plugin marketplace's skills drifts out of sync with the actual code, so remembered problems may already be fixed or wrongly recorded. This Skill closes the loop by re-verifying every remembered claim against live files before changing anything.
Core Features & Use Cases
- Memory-driven audit: Loads MEMORY.md and feedback_*.md files from the per-project memory directory and extracts concrete, checkable claims about skills.
- Verify-before-fix workflow: Classifies each claim as STILL-BROKEN, ALREADY-FIXED, WRONG, or NOT-APPLICABLE by inspecting the actual SKILL.md and plugin.json files.
- Bidirectional correction: Fixes stale skills (promoting soft rules into CRITICAL blocks, adding missing hooks) or corrects stale memory files, then syncs plugin.json and marketplace.json versions.
- Use Case: After weeks of accumulating feedback like "L3 rules get skipped" in project memory, run this Skill to systematically re-check each claim against the current marketplace plugins and apply only the fixes that are still needed.
Quick Start
Ask the assistant to reflect on skills from memory and apply outstanding feedback to the marketplace plugins.