What problem does it solve? Large changes often get committed by file type (models, then services, then tests), producing commits that do not work alone and PRs that exceed reviewer capacity. This Skill structures commits as self-contained work units so each change is reviewable, verifiable, and rollback-safe. ## Core Features & Use Cases - Work-Unit Commit Rules: Enforces committing by deliverable behavior rather than file type, keeping tests and docs in the same commit as the code they cover. - Chained PR Planning: Splits large features into stacked PRs when changes approach the 400-line review threshold, aligned with SDD delivery strategies. - Verification Checklist: Requires focused test results, runtime harness evidence, and explicit rollback boundaries for every commit. - Use Case: When implementing a multi-part feature like token-based authentication, split it into commits such as "add token validation domain model and tests" and "wire token validation into login flow", each independently reviewable and revertible. ## Quick Start Ask the AI to split the current uncommitted changes into reviewable work-unit commits with tests and rollback boundaries included.
