Repository navigation
Wake a watched PR's thread when it merges or closes, and when a review bot reacts #16795
Replies: 3 comments
|
+1, and a few more cases from an audit of ~300 PR-watch runs across my agents (Claude Code, Codex, Grok, Cursor, Feb-Oct). Merge/close and bot reactions are the two biggest gaps for me too. These also force agents back into their own polling loops:
A generic shape that would cover most of this: wake on any change to the PR's head, checks, reviews/comments (created or edited), reactions, mergeability or state, and include the change kinds in the wake. The thread can then ignore what it doesn't care about. Adding an opt-in In my history, PR babysitting logic got reinvented about 47 times and wasted roughly $150 at API prices, mostly working around these gaps. Happy to test a build. |
|
Simple workflow that is annoying without this: agent works, creates PR, there's some back and forth with CI/review, it then merges somehow. I then open the thread and just see some variant of "approved, waiting for merge", whereas what I wanted was "here's what happened, we changed XYZ and reviewer asked for Q, we said no with a reason, reviewer acquiesced and accepted, PR merged @ 14:34 UTC". Sure I could go look at the PR, but maybe the agent didn't link it nicely, I've gotta go find the link in the UI somewhere and anyway who looks at PRs any more lol. similarly, I often have threads who are taking actions as well as writing code, a merge is an important signal for them (e.g. they are waiting for a merge to the kick off a release process, or notify someone via some other channel). |
|
I saw a similar case with a GitHub PR whose only check was optional (
The PR was still open during the wait. I don't know when T3 first had the final check result or whether the missing follow-up and the earlier |
Uh oh!
There was an error while loading. Please reload this page.
Before submitting
Area
apps/server (the
watch_pull_requestreactor).Problem
Two waits come up on almost every PR an agent merges, and
watch_pull_requestcovers neither:gh pr merge --squash --autoand needs to confirm the merge before it reports or starts the next step. The watch ends silently when the PR merges, so the agent writes its ownuntil gh pr view --json state …; sleeploop./issues/<n>/reactionsuntil 👍 shows up or it gives up.On T3 Code 0.0.46-nightly.20261007.2761, across 196 watch wakes from 3 to 7 Oct, I saw none for a merge or a reaction. From 5 to 7 Oct, my agents wrote 23 polling loops for these two waits (16 merge, 7 Codex), 18 of them in threads that had already called
watch_pull_request. The T3 instructions also tell agents not to run their own watcher, so they get mixed signals.Proposed solution
<sha>" or "PR closed without merging".I'm happy to test a build.
All reactions