What problem does it solve? After an AI agent finishes editing code, the changes sit uncommitted in the working directory, often mixed with unrelated dirty files. This Skill identifies exactly which files the agent touched, moves them onto a feature branch via a git worktree, groups them into atomic commits, and opens a pull request without disturbing the user's current branch. ## Core Features & Use Cases - Change Attribution: Cross-references conversation context, Cursor agent transcripts, and git status so only agent-owned changes are committed, never pre-existing dirty files. - Worktree Isolation: Creates a separate git worktree for the feature branch, keeping the user's working directory on its original branch throughout. - Logical Commit Grouping: Splits changes by cohesive concern into conventional commits (feat/fix/refactor/chore) with no AI references in messages. - PR Automation: Pushes the branch and creates a GitHub PR with summary, commit list, and test plan, or appends to an existing open PR. - Use Case: After a long refactoring session, say "commit my changes" and get a clean feature branch with three atomic commits and a ready-to-review PR, while your local checkout stays untouched on main. ## Quick Start Ask the agent to commit my changes and create a PR from this session.
