What problem does it solve? Deciding which parts of a codebase are worth refactoring is often guesswork based on intuition or raw complexity numbers. This Skill produces an evidence-backed, ranked triage of refactor and test-debt candidates using immutable git measurements, so teams can decide WHAT to change before committing to HOW. ## Core Features & Use Cases - Deterministic discovery: Scans a git repository at a frozen commit SHA, combining cyclomatic complexity, fix-commit churn, and configurable criticality weights into a ranked candidate list saved as JSON and a Markdown report. - G1–G5 validation gates: Each candidate is checked for usage/intent, availability (PR and worktree collisions), cohesion, test verification, and bounded change scope, producing decision states like READY, TEST_FIRST, OBSERVE, BUSY, or DELETE_CANDIDATE. - Farm offloading and CodeSift integration: Large repositories can be measured on a remote farm via prepared snapshots, and verified CodeSift envelopes can replace the builtin complexity engine. - Use Case: Point the Skill at a monorepo before a refactoring sprint; it saves a discovery report ranking module families by confirmed pain and criticality, with concrete seams, invariants, and test pointers for each handoff card. ## Quick Start Ask the assistant to run refactor-radar discovery on the current repository and save a ranked report of the top refactor candidates.
