What problem does it solve? Reusable helpers, constants, and test fixtures often end up buried inside workflow-specific packages instead of a shared library, creating duplicate implementations, layer inversions, and maintenance drift. This Skill audits pull requests and packages for those architectural offenses and applies prioritized extraction fixes in small, tested commits. ## Core Features & Use Cases - Offense Detection: Greps PR diffs for ten classified offense patterns (thin wrappers, layer inversions, parallel stacks, duplicate constants) and assigns P0–P3 priorities using a documented taxonomy. - Guided Extraction Workflow: Moves behavior into canonical shared_utils modules, rewires call sites, deletes old modules, and migrates tests alongside the code. - Mode Routing: Supports preflight-proposal, audit-only, and normal fix modes so reviews can be report-only or produce applied changes. - Use Case: A reviewer points at PR #1965 and asks for an extraction audit; the Skill scopes the diff, flags a _default_db_flipper wrapper as P1, extracts it into shared_utils, and runs scoped pytest as evidence. ## Quick Start Ask the assistant to run an extraction audit on PR #1965 and fix any helpers that belong in shared_utils.
