This project welcomes contributions that use AI tools (Copilot, ChatGPT, Claude, Cursor, etc.). Many CNCF projects have adopted similar policies to balance productivity with code quality.
Using AI to help write code is fine. Submitting AI output without review is not.
-
Human verification is mandatory. Every line of AI-generated or AI-assisted code must be reviewed, understood, and tested by the contributor before submitting. You are responsible for what you commit.
-
Run checks locally. AI tools frequently produce code that looks correct but fails linting, formatting, or tests. Before pushing:
make test # unit tests make lint # golangci-lint
-
Match the project's code style. Our style is:
- Terse, single-line comments
- No multi-paragraph godoc on internal functions
- Minimal explanatory comments — the code should speak for itself
-
Don't let AI expand scope. If you're fixing a bug, fix that bug. AI tools often suggest "improvements" to adjacent code. Resist. One logical change per PR.
-
Disclose significant AI usage. If a substantial portion of your PR was generated by AI, note it in the PR description. This is not a penalty — it helps reviewers focus on areas that need extra scrutiny.
-
Sign off your commits. AI cannot sign the DCO for you. Every commit must have your
Signed-off-byline (git commit -s). By signing off, you certify that you wrote or have the right to submit the code.