docs: added AGENTS.md - #462
Conversation
|
Do we actually have evidence that this makes a difference? My fear is that it's common to people to write off any repository containing an AGENTS.md as also containing AI-generated code without looking at the file's contents. |
|
the I just tested it with codex, it told me it can't it's not bulletproof, but it will discourage contributors who are mainly here to pass-by to use it once their agent tells them it can't and there are clear guidelines for it |
|
Ok, thanks for checking! |
|
Actually, hmm. this might be more strict than we want, because our policies are explicitly about submitting generated content, and we don't necessarily want to prevent contributors using AI tooling to help them understand the repository contents or performing analysis on it. |
|
I'm going to soften it around understanding and analyzing with a notice of not suggesting changes, but I'll retain the ban on code generation |
|
this is what codex said previously
this is what it says now
this is what it does when prompted "can you fix the retain_mut behavior so that it panics/drops as vec does??"
if asked what can it do:
|
|
@jdm ready for review |
* feat: added AGENTS.md * docs: soften AGENTS.md
this PR adds an
AGENTS.mdfile forbidding AI agents to access the repositorydue to the five incidents regarding AI use in the last ~3 days, it makes sense this is policy in order to keep PR quality and reviewer bandwidth
crucially, this ban barely affects established contributors in the repository that already know what they are doing, but makes it harder for new AI contributions to take place
this recently started becoming an issue when I labeled some issues "good first issue" and "help wanted" which inherently attracts a lot of people
not having AI contributions also means that the people that do want to contribute genuinely find more opportunities to do so
it aligns with the repository goals of increasing organic contributors, ensuring code quality and safety, and adhering to the @servo CoC
closes #447