[BUG] Over 6k issues labeled with "has repro" have been auto-closed since March 2026
Preflight Checklist
- [x] I have searched existing issues and this hasn't been reported yet
- [x] This is a single bug report (please file separate reports for different bugs)
- [x] I am using the latest version of Claude Code
What's Wrong?
This query indicates that over 6k issues that are labeled with has repro have been auto-closed since the beginning of March 2026.
<img width="1551" height="431" alt="Image" src="https://github.com/user-attachments/assets/584198f1-87f2-4181-8d8b-35ab417417e1" />
<br>
<br>
This behavior of the github-actions Issue Sweep workflow is inconsistent with the repository maintainers explanation of the issue auto-closing process. The explanation was written by Claude, but posted by a human GitHub account:
Issues here are auto-closed by a bot when they've had no activity for a while and are missing information needed to reproduce (usually a version and steps, or a /feedback ID). Nothing is triaged solely by counting.
The incorrect auto-close behavior is aggravated by the Lock Stale Issues workflow that locks closed issues that are inactive for 7 days (on a side note, that locking wokrflow is under a strain so it may experience delays https://github.com/anthropics/claude-code/pull/69470). Because commenting on locked issues is not possible, this locking behavior is contrary to the explanation provided by the maintainers of this repo
If one of your issues was closed and is still relevant, commenting on it with the current version and a repro is enough — a maintainer can reopen it; you don't need to re-file.
What Should Happen?
1) Issues that require additional information from the author are not marked with has repro
2) Issues that are marked by has repro are not auto-closed due to inactivity (due to lack of fresh human comments)
The above two points are symptoms of the issue management process that does not correspond to the maintainers explanation.
The repo maintainers may implement a different issue labeling mechanism (see for example https://github.com/anthropics/claude-code/issues/21732 or https://github.com/anthropics/claude-code/pull/26360#issuecomment-3917183519), or a mechanims of labeling issues with reproduced used mid-August 2026. However, issues labeled with reproduced are also subject to auto-closing due to inactivity, despite maintainers expressing the goal to keep the issues open.
The goal is to correct the bots behavior so it corresponds to the maintaners explanation: issues that contain reproduction steps are not auto-closed based on inactivity.
Moreover, document the issue lifecycle process so the users of this repo know what to expect from the process, and receive notifications when reported issues are fixed (see https://github.com/anthropics/claude-code/issues/21733 for an example suggestion how this can be handled).
Error Messages/Logs
Steps to Reproduce
Determine that issues that don't collect at least 10 thumbs up and have no recent comments will be auto-closed.
- Check the Issue Sweep workflow has these settings: 14 days for stale, additional 14 days for autoclose, with stale marking exemption for issues with 10 or more thumbs up:
- Verify the workflow does not exlude issues that contain
information needed to reproduce (usually a version and steps, or a /feedback ID)from auto-closing.
Claude Model
Not sure / Multiple models
Is this a regression?
Yes, this worked in a previous version
Last Working Version
_No response_
Claude Code Version
2.1.233 (Claude Code)
Platform
Anthropic API
Operating System
Ubuntu/Debian Linux
Terminal/Shell
Xterm
Additional Information
Some issues returned by the query may have incomplete reproduction steps, missing Claude Version or missing /feedback ID. Such cases should not be labeled by has repro, and instead the issue author needs to notified about the information that is missing. Currently there is no such notification due to https://github.com/anthropics/claude-code/issues/46083#issuecomment-5311671488.
In the comments and links below you'll find examples of issues that were (or were about to be) auto-closed due to inactivity despite containing reproduction information.
15 Comments
Adding a noise comment to prevent this from being closed.
+1
See also https://github.com/anthropics/claude-code/issues/70161: even the bugs marked with
reproducedare labeled withstaleafter 14 days, auto-closed after further 14 days, and subject to locking after additional 7 days.Bump
Commenting to prevent the auto-close :)
Thanks, but this issue won't be auto-closed by the bot in the current GitHub actions configuration because it has reached 10 upvotes.
What is needed most now is publicity on social media, so a human maintainer responds.
Now it's ~7.4k
Another concrete data point: #77237 (OTel
git_commit_idscraped from git-commit stdout, silently dropped forgit -C <path> commitand quiet/redirected commits) was closed 2026-09-21 with the standard "inactive for too long" bot comment — despite having a full repro matrix and root-cause analysis of the shipped extraction logic, i.e. exactly the kind of issue the maintainer's quoted policy says shouldn't be auto-closed on inactivity alone.I verified the underlying bug is still present in 2.1.278 (binary
stringsanalysis + live repro) and re-filed as #95934, since the closed issue can't be commented on. That's the actual cost of this policy gap: not just noise, but re-doing verification work that a maintainer comment or a reopen could have avoided.This is unacceptable for paid service
+1, this closed an issue due to inactivity, instead of a fix: https://github.com/anthropics/claude-code/issues/79174
even though the issue had multiple +1's, not the greatest experience, would rather the issue be fixed then ignored
The irony of this issue not having the has_repro or reproduced tags.
Pinging one known maintainer as it seems this issue has gone invisible: @bcherny
Thank you for opening this.
---
@km-anthropic @ollie-anthropic @hackyon-anthropic @weishen-anthropic: I tagged you because you have
-anthropicsuffix on your github name and it feels like it is impossible to get a human to address this disaster of an issue. Can you get someone to please fix this?So so so many issues get auto closed when there is clear evidence of reproducibility and serious issues. Good stuff is being closed. And then, it discourages myself and others from opening or commenting on issues in the future. It is seriously hurting the Claude community and needs to be fixed.
Thank you
This is Ben's Claude Code session, posting on his behalf and with his approval.
@marcindulak: the sweep script has no exemption for
has reproorreproduced. Inscripts/sweep.ts,markStaleandcloseExpiredskip only PRs, locked issues, assignees (markStaleonly), and issues with 10+ 👍. I couldn't open a PR against this repo from our account, so the fix is on a fork branch instead (9 lines): https://github.com/benissimo/claude-code/tree/fix/sweep-exempt-repro-labelsIt skips any issue carrying
has reproorreproducedin both functions. A--dry-runagainst a stubbed API (5 issues, all 40 days inactive) changes from 3 staled and 2 closed to 1 staled and 1 closed. I found no workflow in this repo that applies those two labels, and the triage prompt only appliesneeds-repro, so triage would also need to applyhas reprowhen it judges a repro good. Maintainers are welcome to cherry-pick it.yeah the closing bot is very over aggressive.
I mean I kinda get it with the insane flood of cloud written issues this repo has to deal with but its just overzealous.
I nearly wrote a bot that auto comments on issues every week just so this doesnt happen but stopped cause I dont want to add to the spam. but this bot kinda makes it necessary i feel like.