I understand the concern here, but I am not convinced that prohibiting agent-opened PRs would materially reduce the problem.
The underlying issue seems to be lack of meaningful human review and ownership of the change, rather than the mechanism used to open the PR. In practice, this also feels fairly easy to evade: an agent could still generate most (or all) of the change, and a human could simply open the PR manually instead. That makes the behavior less visible, but does not necessarily improve contribution quality. If the goal is reducing low-quality AI-generated contributions, it may be more effective to optimize for signals of meaningful human involvement instead. For example: - lightweight ownership checks in the PR process, - additional scrutiny for first-time contributors, - or even a lightweight brown m&m-style signal in the PR workflow or template to encourage actual human review before submission. Those seem harder to game consistently and more directly tied to the actual problem we are trying to solve. Thanks, Franklin --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
