Enforce Responsible Usage of AI

[wu21-web]

Hi, I am wu21-web. I have recently started learning about this project. I previously contributed to hustcer/deepseek‑review, and I have now found a better alternative. I am currently getting familiar with its codebase.

This project hit Hacker News, where I observed many contributors leveraging AI for development (@claude co‑authored more than ten commits). Some contributors submit huge diffs on their very first attempt — a practice widely referred to as “vibe coding”. I fell into this pattern once or twice myself and now regret it. In many pull‑request diffs, tests generated by AI occasionally include dead logic.

For this reason, I believe it is essential that every contributor manually reviews their changes at least once before opening an upstream pull request. Do not misunderstand me: I am not opposed to using AI/LLMs — this project itself centers on that very topic. My concern lies in responsible usage.

I have also noticed contributors posting long, detailed comments on issues and pull requests. This practice is valuable for debugging, yet repetitive “xx‑xx” phrasing patterns sometimes make me wonder whether I am conversing with an AI.

I propose adding guidelines covering responsible AI usage to both our Code of Conduct and contributing documentation. I greatly admire the approach taken by @MikeMcQuaid on this subject. Below are draft guideline items:
A. Always review AI-generated changes (at minimum the core logic) prior to submission.
B. Do not request a maintainer review until you have reviewed your own changes.
C. Pull requests that are AI-assisted or AI-authored (excluding draft PRs) must be explicitly marked as such within their descriptions.
D. You may use AI to aid your review of production-ready changes, but a full manual review remains mandatory.
E. Avoid non-agentic AI tools when drafting submissions or performing reviews.
F. You may not list an AI/LLM as a commit author, co-author, committer or signatory. This prohibition applies to commit trailers including  Assisted-by ,  Co-developed-by  and similar tags.
G. You must personally respond to all questions and review comments from maintainers, without relying on AI/LLMs to compose your replies.

To conclude:
I do not seek to ban AI assisted contributions. The bare minimum expectation is that you review AI
generated work yourself to guarantee correctness and foster a trustworthy community.

Happy Coding,
wu21-web

P.S. If anyone adapts this text for inclusion in CODE_OF_CONDUCT and the contributing guide on my behalf, please, please do not use AI.