What problem does it solve? Completed task feedback and PR review threads accumulate valuable lessons, but they stay scattered across files and are never turned into actionable guidance. This Skill mines that history and converts repeated observations into durable workflow rules. ## Core Features & Use Cases - Feedback Scanning: Reads all docs/tasks/*/feedback.md files and merged PR review threads, tracking processed files to avoid re-scanning. - Theme Clustering with a Strict Threshold: Groups observations by theme and only promotes a theme to a rule when it appears in 3 or more independent task examples; weaker themes are reported but dropped. - Rule Authoring and Maintenance: Drafts rules in imperative form with use_when triggers and verbatim evidence, then merges them into docs/rules.md with sequential IDs and commits the result. - Use Case: After completing a dozen tasks, run the analysis to discover that missing tenant isolation appeared in four separate reviews, and codify it as a rule so future planning catches it. ## Quick Start Ask the agent to analyse feedback across completed tasks and extract new rules into docs/rules.md.
