In contributions

AI-assisted contributions are welcome. What matters is that the contribution is genuinely yours: understood, checked, and defensible in review.

If you are contributing

You own what you submit. Whatever wrote the first draft, you are the author. Be ready to explain any line in review — how it works, why it is correct, what happens on bad input.
Say how AI was used. A line or two in the pull request description is enough: “Used an assistant to draft the test fixtures; the parser changes are mine.” Reviews teach better when the reviewer knows which reasoning was yours.
Don’t submit what you haven’t read. A pull request generated wholesale by an agent, which its author cannot walk through, will be sent back.
Check dependencies and security-relevant code yourself. Confirm every new package exists and is the one you meant. Give extra scrutiny to input handling, authentication, file access and queries.
Keep private material out of AI tools. No credentials, tokens, private repository content, or anything that must stay out of a public repository.

If the contribution is how you are learning

When the point of the work is to build a skill — a first issue, a mentored task — follow While you’re learning: write the core logic yourself, and use AI to understand rather than to produce. Getting stuck, and then asking a person, is part of the process, not a failure of it.

If you are reviewing

Review AI-assisted code at least as hard as any other. The research shows AI raises confidence faster than it raises correctness.
Ask the author to walk you through it. Not as a penalty — it is the review doing its job, and the quickest way to find an “almost right” section.
Check that new dependencies exist and are the packages intended.
Look for duplication where existing code should have been reused.
Keep it a lesson. See Reviewing.