jameslamb's Description of Work
I've been doing the majority of the issue triage and PR review in this repo over the last year, and in that time I've seen a lot of the same problematic things many other open source maintainers have been experiencing with the mass adoption of AI coding agents.
- low-effort PRs from new contributors who just told an AI system "open a PR that fixes
<issue link>" - AI-generated PRs trying to compromise the project's security
- contributors suddenly closing PRs and disappearing in response to a thoughtful review I spent significant time on
- contributors pasting raw LLM output back at me without clearly stating that
This PR proposes formalizing some of the policies I personally have been applying into something written in LightGBM's documentation. Hoping that that will reduce the effort to deal with such things (can just link to a policy), and that this PR can be a place where we (LightGBM maintainers) can come to a consensus about how we all want to approach these things.
Notes for Reviewers
How I tested this
Enabled readthedocs builds on this branch.
docs: https://lightgbm.readthedocs.io/en/docs-automated-contribution-policy/CONTRIBUTING.html