| English | 中文 |
This directory discusses Git workflows in the age of AI programming: AI can generate code faster, and humans need to use Git to organize changes into reviewable, verifiable, and rollback-ready engineering units.
It belongs to the “complete an engineering change” layer of the knowledge map. First distinguish the working tree, index, and committed revision; when you need to choose a team collaboration model, continue to Team Collaboration.
Start with the engineering change course. Connect the task contract, commit scope, CI, runtime evidence, and recovery; then use Accept an Agent Change for evidence fields and boundaries.
The companion local artifact-and-service exercise builds an artifact in a temporary Git repository and starts a minimal HTTP service only on loopback to check a runtime result. It does not contact real CI, a hosting platform, production, or an external service. The command-safety regression covers only the four documented command patterns; remote authorization, branch protection, hooks, and team rules still need checks in the actual repository.
| Problem you want to solve | Recommended Reading |
|---|---|
| Establishing a Git workflow for AI programming | AI Native Git Workflow |
| Taking a small repair from task contract through recovery | Engineering Change Course |
| Don’t know how to review code after AI modifications | AI Change Review Practice Example |
| Making CI name the actual candidate or integration revision | CI for AI-Generated Changes |
| Systematically reviewing AI-generated code | How to Review AI-Generated Code |
| AI modified too many files at once | AI Commit Splitting |
| Large diffs need to be split into multiple PRs | Stacked PR for AI-Generated Changes |
| Parallel development with multiple agents | Multi-Agent Branch Strategy |
| Reclaiming a background agent result | Background Agent Tasks |
| Assessing whether Agent adoption improves collaboration | AI Collaboration Metrics |
Tools change, but underlying principles remain stable: isolate tasks, review diffs, split commits, retain validation, and let humans be responsible for the final merge.
| Scenario | Recommended Combination |
|---|---|
| Single small fix | AI Workflow + AI Review |
| Large feature | AI Workflow + Commit Splitting + Stacked PR |
| Multi-agent parallel work | Worktree + Multi-Agent Branch Strategy |
| Team adoption of AI Review | AI Generated Code Review + AI Reviewer and Human Reviewer |
| Background agent work | Background Agent Tasks + CI for AI-Generated Changes + AI Agent Incident Recovery |