What problem does it solve?
Large changes often land as monolithic commits that reviewers cannot audit, bisect, or revert cleanly. This Skill guides you to split work into standalone commits that each compile, pass focused tests, and make sense on their own at their point in branch history.
Core Features & Use Cases
- Commit Boundary Checks: Six acceptance checks (reviewable, buildable, complete, focused, auditable, revertible) validate each staged diff before committing.
- Dependency-Ordered Waves: A wave pattern (foundation, factories, contracts, infrastructure, consumers) orders multi-file changes so every commit compiles.
- Split/Combine Heuristics and Anti-Patterns: Concrete rules for when to split refactors from behavior changes and when to combine types with their implementations, plus examples of teaser commits, hidden behavior, and cleanup hitchhikers.
- Use Case: When refactoring a schema across types, factories, contracts, and app code, plan five waves, stage each wave's files, verify with a type-check, and commit each as one coherent unit.
Quick Start
Ask the AI to review your staged changes and split them into standalone, dependency-ordered commits using the standalone-commits acceptance checks.