What problem does it solve? After running a refactoring audit, teams have a list of findings but no safe order to execute them. This Skill turns findings registers into a sequenced plan where every step is small, behaviour-preserving, independently landable, and annotated with risk, rollback, and test coverage — while routing oversized work to the task pipeline instead of pretending it fits in one commit. ## Core Features & Use Cases - Findings re-verification: Re-checks each finding's file:line evidence against the current HEAD before planning, dropping stale or already-fixed items. - Size routing: Classifies findings as inline-plannable small steps or routes them to the task pipeline with a ready-to-use task-start input block. - Dependency-first sequencing: Orders steps by dependency then risk, with every step carrying rationale, rollback, tests, and effort fields. - Use Case: After running refactor-audit on a legacy module, ask for a plan covering findings F1 and F3; the Skill writes a plan file under .marvin/refactor/ with ordered steps and hands oversized items to the spec pipeline. ## Quick Start Ask the assistant to plan the refactoring for findings F1 and F3 from the latest audit report under .marvin/refactor/.
