What problem does it solve? Large-scale code cleanup — removing dead code, unused imports, and duplication across many files — is risky to do by hand and slow to do serially. This Skill automates the whole pass: it chunks the scope by import graph, runs parallel optimizer agents, verifies tests and linters after every batch, and loops reviews until every chunk passes. ## Core Features & Use Cases - Import-graph chunking with blast-radius analysis: resolves a glob, directory, git range, or free-text scope into parallel-safe chunks, isolating high-risk files and surfacing downstream impact before any code is touched. - Parallel optimize-verify-review pipeline: dispatches up to 6 optimizer agents with mirrored code reviewers, auto-fixes regressions in an inner loop, and respawns only rejected chunks. - Behavior preservation guarantees: captures a test/lint baseline first, requires explicit user approval at a plan gate, and never proceeds on a red build. - Use Case: Run it on "src/auth/" after a refactor sprint — it removes dead exports and unused imports across the module, keeps tests green, and delivers a summary of every change plus deferred observations. ## Quick Start Ask the assistant to clean up dead code and reduce complexity in the src/auth directory using the cleanup skill.
