What problem does it solve? AI-assisted coding sessions often end with a pile of uncommitted changes dumped into a single vague commit, or worse, committed directly to main. This Skill enforces disciplined version control: one logical change per commit, conventional commit messages, feature-branch isolation, and a gated push-and-PR workflow. ## Core Features & Use Cases - Commit Mode: Surveys the diff, groups changes into logical units, and stages each unit separately (including partial staging with git add -p) to produce clean atomic commits with conventional messages. - Ship Mode: Extends Commit mode with base-branch resolution, rebase, a four-part pre-push gate (branch safety, base verification, quality gates, secret scan), push, and PR creation via the gh CLI. - Copilot Review Request: Automatically requests a Copilot pull request review after PR creation, with REST API verification and retry fallbacks. - Use Case: After finishing a multi-file feature, ask the agent to "ship this work" — it creates an ai/feat/... branch, splits the diff into atomic commits, rebases onto the verified base branch, runs tests and a secret scan, pushes, and opens a PR ready for review. ## Quick Start Ask the agent to commit my current changes as atomic commits, or say "ship this" to also push and open a pull request.