flag pull requests likely to be fully generated by an LLM - #20148
flag pull requests likely to be fully generated by an LLM#20148zacharyburnett wants to merge 3 commits into
Conversation
|
Thank you for your contribution to Astropy! 🌌 This checklist is meant to remind the package maintainers who will review this pull request of some common things to look for.
|
| close-pr: false | ||
| lock-pr: false |
There was a problem hiding this comment.
with auto-close and auto-lock set to False this will just make a comment on the PR
There was a problem hiding this comment.
good strat as a starting point. I hope we can move to auto-close later if this proves reliable and useful.
|
While I applaud the enthusiasm, I am hesitant to introduce an experimental workflow (v0) from an author we know nothing of. Is there another repo we can try this on that is not a core library? |
|
sure, but we might have to test it ourself with our own agent, since we're not currently getting a huge amount of slop PRs in. Maybe that's a good thing though, and we can wait until it's a bigger problem to revisit this |
|
Surely more popular upstream packages are in more desperate condition. We could see which anti-bot bot they use and go from there. |
| pull-requests: write | ||
|
|
||
| on: | ||
| pull_request_target: |
There was a problem hiding this comment.
as pointed by zizmor: this trigger is really a security can-of-worm. I'm surprised this is actually from a documented example.
There was a problem hiding this comment.
@zacharyburnett how do we know that simply changing this to pull_request still works as intended ?
There was a problem hiding this comment.
as far as I understand it, pull_request_target just runs the workflow from the target branch, so changing it to pull_request should still trigger the workflow normally (from the PR branch). There aren't any checks in the anti-slop workflow itself for github.event_name or pull_request_target so it shouldn't matter what the trigger is
There was a problem hiding this comment.
I'm not convinced yet. I opened an issue to touch base with the author: peakoss/anti-slop#20
Description
adds https://github.com/peakoss/anti-slop workflow, perhaps it will be useful for reviewers? We are also interested in knowing if it will work correctly @pllim