# xirong/my-git: A Git Change Control Handbook for AI-Native Engineering

> The repository is a Chinese-language handbook and interactive lab set that treats Git as the control plane for reviewing, constraining and recovering AI-generated code changes. It is documentation, not a tool, and the MIT licence makes reuse easy.

**xirong/my-git** — Git as the control plane for AI-native software engineering | AI Native 软件工程的 Git 变更控制手册

- Repository: https://github.com/xirong/my-git
- Website: https://github.com/xirong/my-git
- Stars: 7,399 · Forks: 2,477
- Language: HTML
- License: MIT
- Published: 2026-09-22 · Updated: 2026-09-22 · Language: en
- Canonical page: https://hysenlabs.com/projects/xirong-my-git

## What xirong/my-git is, and who it is written for

This is a handbook, not a library. The repository is an HTML-heavy documentation set with a Chinese README and an English counterpart, organised into numbered directories from 01-getting-started through 10-company-practices. The stated audience is broad: Git beginners building a mental model, working developers who deal with conflicts and rollbacks, senior developers who care about pull requests, review and CI, and tech leads who design team workflows and GitHub governance.

The distinguishing part is the last audience in that list. The README names AI coding users who need to control the code-change risk introduced by Codex, Claude Code, Cursor and Cline. That framing drives a whole directory, 05-ai-native-development, with pages on accepting agent changes, CI for AI-generated changes, background agent tasks, worktree use for agents, stacked PRs for generated changes, and AI collaboration metrics. If you are trying to decide whether to adopt this repository, the question is not whether it runs, because there is nothing to run. The question is whether its written positions on reviewing machine-written diffs match how your team already works.

## The mechanism: a knowledge map that routes you to one of three problems

The README does not present the material as a linear book. It offers a knowledge map and asks the reader to first decide whether the problem in front of them is a principle, a single engineering change, or team collaboration. Each answer points at a different entry page. Principles lead to the interactive mental-model lab and a full learning path. A single change leads to a page called the complete acceptance of one change, plus a worked AI change review example. Collaboration leads to the team Git workflow guide and the GitHub engineering governance handbook.

That routing is the real architecture of the project. The content is grouped so that a reader who is confused about where a change lives, how history forms, and what recovery depends on gets sent to the interactive material, while a reader who needs to judge whether scope, commit version and test evidence correspond gets sent to the change-control pages. The README is explicit that Git alone is not the whole story: testing, review, permissions, artefacts and release systems together form complete change control. The handbook claims to describe both what the tools can do and where their limits are.

## Installing nothing: how to read the handbook and run the interactive labs

There is no package to install. The README points readers at the published interactive site and at the repository's own directories. If you want the material locally, the only step the repository supports is cloning it, and then opening the Markdown or the interactive HTML. The README gives no npm package name, no build command and no server port, so nothing of that kind belongs here.

```bash
git clone https://github.com/xirong/my-git.git
cd my-git
```

The README states that the interactive learning experience is published at a GitHub Pages URL for the mental-model lab, and that the ten topics are arranged in the same order, with 01 to 03 corresponding to existing interactions and 04 to 10 to newly added ones. It also states that the material and scripts can be inspected, and that reader understanding still needs to be verified through prediction, reproduction and transfer. That is a fair description of what you get: articles, an interactive page, and lab scripts in the labs and scripts directories, not a graded course.

```bash
ls 01-getting-started 05-ai-native-development 06-troubleshooting
```

After cloning, those three directories are the ones a new reader is most likely to open first. The README's own first-stage reading list points at the AI Native Git Workflow page, the Codex and Claude Code practices page, the AI coding tool Git integration page, and the AI change review example.

## Where the handbook stops being enough

The repository describes conventions and review discipline. It does not enforce them. A page on branch protection or rulesets explains how to configure GitHub, but the configuration lives in your repository settings, not in this one. A page on agent governance explains what an AGENTS.md file should say, and the repository ships a template in 08-templates, but nothing checks that an agent obeyed it. If your problem is that agents push directly to main, reading the governance page will not stop them. A ruleset or a required review will.

The language boundary is the second constraint. The primary README is Chinese, with an English file alongside it. The repository's primary language is listed as HTML, which reflects the interactive material rather than the prose. A team that needs an English-first reference has one README and the linked English files, and should confirm what the English README actually covers before standardising on it. The README also notes that Git is only the version and history foundation, and that testing, review, permissions, artefacts and release systems complete the picture. That is an honest boundary, and it means the handbook cannot be the only document in your change-control process.

## How it differs from Oh My Git and from a plain Git tutorial

The related searches around this repository include Oh My Git, which is a different project: an interactive game for learning Git commands and the commit graph. The overlap is the interactive teaching approach. The difference is scope. Oh My Git teaches the command layer through play. xirong/my-git teaches the design layer through articles and labs, then extends into review, governance and AI agent collaboration. If your team already knows the commands and the problem is that generated diffs arrive faster than reviewers can read them, the game format does not address that, and this handbook does, at least in writing.

Against a conventional Git tutorial, the split is similar. Most tutorials stop at rebase and stash. This repository keeps going into CODEOWNERS, merge queues, release management, partial clone and sparse checkout for large repositories, secret removal from history, and recovery after a force push. The 06-troubleshooting directory includes a playbook, an undo-anything page, and a page specifically about recovering from an agent incident. That last page is the clearest signal of what the maintainer thinks the new failure mode looks like.

## Maintenance, licence and what upgrading costs

The repository is not archived, and the last push was on 2026-09-09. Releases are versioned and dated: v2.0.1 on 2026-06-07 covered open source governance and maintenance updates, v2.1.0 on 2026-06-11 added AI agent governance and conventions, and v2.2.0 on 2026-09-09 added Git learning labs and AI collaboration practices. The cadence suggests the maintainer is still adding material rather than only fixing typos.

Upgrade cost is low in the software sense, because there is no dependency to bump. The real cost is editorial: the AI-native pages describe practices around fast-moving tools, and a team that pins its internal guidance to a specific page should expect that page to change between minor versions. The licence is MIT, which permits reuse and modification with the licence and copyright notice retained. That matters if you want to fork the templates into an internal handbook. It is not legal advice, and if you plan to redistribute modified content, read the LICENSE file in the repository root.

## Conclusion

Adopt xirong/my-git if your team is already letting Codex, Claude Code, Cursor or Cline write code and you need a shared vocabulary for reviewing, splitting and recovering those changes. Skip it if you want a runnable governance product, an English-first text, or a replacement for your CI and branch-protection configuration. Before relying on it, verify the English README, the AGENTS.md template and the AI change control loop page, since those are the parts a non-Chinese-speaking team would actually lean on.

## FAQ

### Is xirong/my-git a tool I install, or documentation?

It is documentation. The repository is a handbook with articles, interactive pages and lab scripts, and the README gives no package to install or service to run. The only setup step it supports is cloning the repository and opening the Markdown or the interactive HTML.

### Does xirong/my-git cover AI agent governance?

Yes. The repository has a page on AI agent governance under 04-github-engineering, and the v2.1.0 release was titled AI Agent Governance and Conventions. It also ships an AGENTS.md template in 08-templates for teams writing their own agent collaboration rules.

### What licence does xirong/my-git use?

The repository is MIT licensed. That allows reuse and modification provided the licence and copyright notice are kept, which is relevant if you want to fork the templates into an internal handbook.

### Is the handbook available in English?

There is an English README, README_en.md, linked from the top of the Chinese README. The primary documentation set is written in Chinese, so a team that needs English-first material should check the English file before standardising on it.

## Sources

- [License: MIT](https://github.com/xirong/my-git/blob/master/LICENSE)
- [Project website](https://github.com/xirong/my-git)
- [README](https://github.com/xirong/my-git/blob/master/README.md)
- [Releases](https://github.com/xirong/my-git/releases)
- [xirong/my-git on GitHub](https://github.com/xirong/my-git)

---

Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/xirong-my-git