What problem does it solve? Before committing code, developers need to run type checks, linters, builds, and tests — but every repo uses a different toolchain, and fixing the mechanical failures they surface is repetitive manual work. This Skill detects the project's own toolchain and runs a self-healing local quality loop that fixes what it safely can. ## Core Features & Use Cases - Toolchain Detection: Reads CLAUDE.md/AGENTS.md, turbo.json, package.json scripts, Makefile, pyproject.toml, Cargo.toml, or go.mod to run the repo's own commands — never hardcodes a stack. - Self-Healing Loop: Runs static analysis in up to 5 iterations, auto-fixing safe issues (imports, type mismatches, formatting, debug residue) while flagging complex problems for human review. - Deslop Scan: Removes debug output, commented-out code, unexplained suppressions, and escape-hatch types from the changed diff only. - Use Case: After finishing a feature in a TypeScript monorepo, ask the agent to check your uncommitted changes — it runs the Turbo pipeline, fixes lint and type errors, verifies build and tests pass, then hands off to smart-commit with --commit. ## Quick Start Ask the agent to check my uncommitted changes and fix any lint, type, or test failures before I commit.