[BUG] github-actions incorrectly closed a bunch of issues for "60 days of inactivity"

@catherinewu looks like github-bot also ignored you on #3039 :(

Hopefully @ashwin-ant can comment on this behavior.

Updated my original comment so the discussion is more clear: There was some recent activity around this https://github.com/anthropics/claude-code/commit/5e3e9408feea99c66bd2344f0e19b5f25d2a6b11, maybe the removal of autoclose was failing before this revert?

Actually, on the examples I pointed out the autoclose label was correctly removed, but the issue was still autoclosed :(

I see, let's also tag @dhollman, who comitted the "60 days of inactivity" workflow https://github.com/anthropics/claude-code/pull/6369, and @bcherny and @km-anthropic, who reviewed it.

Interestingly, the above PR, was at least party "Generated with [Claude Code]", and its description contains this comment "Closes issues that have been inactive for 60 days (where last comment was the warning)". This requirement could be treated by the machine that human activities are not taken into account for "60 days of inactivity", or misimplemented.

In the meantime you can upvote all discussions listed in https://github.com/orgs/community/discussions/182939. If the author of the issue could reopen it, the damage by autoclose bots would be smaller (similar damage is being/was made to other repos, e.g. https://github.com/Azure/AKS/issues/3474).

I’ve had this happen to so many issues. Very frustrating.

@marcindulak - thank you for linking me to this issue.

Thank you @marcindulak for bringing this to my attention, and to @ANogin for initially reporting it.

Even when functioning properly, auto-close bots are harmful to community contributions:

• Auto-closing without proper resolution is disrespectful to the original reporter and anyone else who contributed to the report.
• It just sweeps the issue under the carpet, blindly hoping that it will magically fix itself. (The fact that this can sometimes happen doesn't justify doing it in all the other cases.)
• When issues or feature requests are marked as closed without resolution, then people are more likely to miss them, resulting in duplicate reports and consequently fragmented information.
• People searching for ways to contribute are much more likely to miss closed issues too.
• Long-standing issues are less likely to get resolved. Here's an example of a success story when issues can remain open for a long time.

See also this post.

I do understand it is annoying and perhaps even stressful to have open issues going nowhere. One solution for this is to label short-lived and long-lived issues differently to distinguish them. Another is assigning issues to different milestones. This makes it clear to users which issues are likely to be fixed sooner rather than later. Proper triage processes are valuable, but simply closing issues is not the solution. Personally I have found this highly demotivating, and will reconsider whether it's worth contributing feedback (for free, as a paying customer) to Claude Code in future, given that it generally just gets ignored.

this is very annoying, please fix this. Your Claudes are taking over, Anthropic

@marcindulak - I've noticed issues are now also being locked due to having no activity since being closed, despite you having commented on them since they were closed (to link to this issue).

For example: https://github.com/anthropics/claude-code/issues/7760#issuecomment-3768564443

I think one would need to comment repeatedly every week to prevent the "has not had any activity for 7 days" bot from locking the issue.

Let me check this assumption by commenting on https://github.com/anthropics/claude-code/issues/8528#issuecomment-3773927503

@bcherny this is a big issue for anthropic systemically getting the feedback for active and ongoing Claude Code bugs. You might want to take a look

https://github.com/anthropics/claude-code/issues/11804

I think one would need to comment repeatedly every week to prevent the "has not had any activity for 7 days" bot from locking the issue. Let me check this assumption by commenting on #8528 (comment)

There is an interesting outcome of this experiment:

1. Issue with a fresh comment - not locked https://github.com/anthropics/claude-code/issues/8528#issuecomment-3773927503
2. Issue without a fresh comment - claimed to be bo locked, but not locked https://github.com/anthropics/claude-code/issues/9675#issuecomment-3792145177
3. Issue without a fresh comment - locked https://github.com/anthropics/claude-code/issues/11009#issuecomment-3784635761

I won't be opening any issue about this locking failure (Failed to lock issue #9675: Requires authentication) problem yet.

<img width="1101" height="299" alt="Image" src="https://github.com/user-attachments/assets/142a6c0c-c42d-44b0-bb30-ac5068bfb94d" />

Fix here if anyone wants to contribute a review.

The bot is now locking those same issues, such as https://github.com/anthropics/claude-code/issues/516. This is a problem.

Locking the issue after 7 days of inactivity after the issue is closed works as designed, except the locking bot fails sometimes to lock. For details see https://github.com/anthropics/claude-code/issues/16497#issuecomment-3792158713

so much talk about AGI but we can't get a GitHub bot to work correctly

For your information, there are at least two more broken github action workflows, in addition to our "60 days of inactivity" one:

https://github.com/anthropics/claude-code/issues/19267 [Bug] The github bot creates impenetrable webs of duplicate tags in bug reports -- all of which will be closed in 3 days
https://github.com/anthropics/claude-code/issues/21953 [BUG] Oncall Issue Triage fails to find relevant issues

#6213 is still an issue for me (has been since August) and was just incorrectly closed.

100th 👍!

Example I came here from: https://github.com/anthropics/claude-code/issues/1317

reminder: FIX is open for review:

<img width="3113" height="3045" alt="Image" src="https://github.com/user-attachments/assets/13090a09-26af-41c3-a1b6-49286f304aac" />

@marcindulak please tell me you're using either @github or @claude to at least semi-automate that. Strong work. I approve of your effort!

The workflow that closes issues due to "60 days of inactivity" is here https://github.com/anthropics/claude-code/actions/workflows/stale-issue-manager.yml. and it prints what issues it closes.

I'm not collecting the issues that were incorrectly closed due to "60 days of inactivity", but that from the human comments appear to be resolved. There is currently no way to reliably automate this decision step. It would require a human product manager familiar with Claude Code behavior on various platforms to look at every issue https://github.com/anthropics/claude-code/issues/21732.

Moreover, fully automating issue triage would remove some of the laughs from generative AI’s failure modes.
One of my favorites until now is https://github.com/anthropics/claude-code/issues/10580.

16497 also appears to have been closed incorrectly. It is now locked so I can't add a comment there.

Blocking user comments on incorrectly closed issues seems like this should be a fire alarm

#11282

Tangentially, I notice that https://github.com/anthropics/claude-code/issues/22578 was closed as a duplicate even though the instructions say it shouldn't be if the duplicate comment was downvoted. (It's possible that someone downvoted _after_ it was closed, but that seems unlikely.)

(Even more tangentially, I guess that issue really is a duplicate; but the bot should follow the instructions it posts.)

Tangentially, I notice that #22578 was closed as a duplicate even though the instructions say it shouldn't be if the duplicate comment was downvoted. (It's possible that someone downvoted _after_ it was closed, but that seems unlikely.) (Even more tangentially, I guess that issue really is a duplicate; but the bot should follow the instructions it posts.)

Yes, I also noticed that downvoting is (sometimes?) insufficient to prevent the closure as a duplicate https://github.com/anthropics/claude-code/issues/19267#issuecomment-3777595010

I can already see the next OpenAI ad where a client tries to reach customer service and receives "your problem was auto-resolved, thank you for contacting us".

I like Anthropic products, but especially as a company that heavily targets a software dev audience, this is not the message you want to send to your customer base.

@ant-kurt sorry to ping you personally but what would it take to get Anthropic to acknowledge this problem?

I've stopped filing GitHub issues and sharing feedback for Claude Code because of this issue with the bot closing issues + the lack of human support here.

This is really bad customer experience and it's one of the reasons I've started using Codex CLI more often as of lately. Their GitHub issues channel is curated by humans.

@PaulRBerg yes, exactly. The only reason I even saw this issue was because an issue I opened on Claude Code 8 months ago was closed despite recent human comments. I switched to Codex about a week after I opened that issue and haven't looked back. Wasn't surprised to see that that issue hadn't received dev attention at any point.

Hey yall I'm not gonna lie I would love to get a PR merged into this repo, but I'd also like to suggest that reviewing an open PR fix for this issue might be a sign of good faith that is something important to you to get fixed and might make it more likely that someone from the team would consider reviewing it. 🙏🏽

@fvolcic @hackyon-anthropic no response from ant-kurt so trying the next couple of non-bot accounts in the commit history. Could you get somebody at Anthropic to take action here?

For your information: if you update the issue by posting a comment around the time the oncall labeller bot runs https://github.com/anthropics/claude-code/issues/21953 it may consider the issue for "oncall" https://github.com/anthropics/claude-code/actions/runs/21864553305. However, the AI thinks that this issue is unimportant. It does not understand that the misconfigured "60 days of inactivity" bot results in active issues being closed.

GitHub Actions incorrectly closing issues (repository automation, not a product bug)

<img width="1562" height="303" alt="Image" src="https://github.com/user-attachments/assets/f737b1c9-e6e8-4583-94a2-b34e3e0ba965" />

Morever, the "oncall" label does not make the issue automatically noticed by a human, but that's a different topic https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20state%3Aopen%20label%3Aoncall

Youre doing the lords work @marcindulak : D

It starts locking issues now https://github.com/anthropics/claude-code/issues/10895#issuecomment-3840418356

It starts locking issues now #10895 (comment)

Locking of closed issues after 7 days of inactivity works as designed, see https://github.com/anthropics/claude-code/issues/16497#issuecomment-3792158713

My accessibility issue (#13074) was one of the ones incorrectly closed by this bot. I had responded to the 30-day warning confirming the issue was still active, included a detailed root cause analysis showing the problem requires changes within Claude Code itself, and the bot closed it anyway. It's now locked so I can't reopen it.

This is especially frustrating for accessibility-related issues. Blind and screen reader users already face higher barriers to reporting bugs—the process of writing detailed reproduction steps, testing workarounds, and documenting root causes takes significant effort. Having that work silently discarded by an automated process, with no human ever acknowledging it, is demoralizing.

I'm a paying Claude Code user. I would rather have my issue sit open and unresolved than have it closed and locked with no path to follow up.

Auto-closing issues causes duplicate issues to be created, so you then you lose the ability to count how many 👍's an issue has to determine how many people it affects. Instead you have to track down all of its duplicates and sum them.

Auto-locking issues prevents users from being able to leave comments that help each other. Unless some very specific kind of abuse is being prevented by locking issues, it should almost never be done.

I would suggest that Anthropic reconfigure these bots to apply/remove labels instead.

Honeslty I'm done - just had the issue I care about closed and blocked from reopening. Between this, treatment of openclaw, opencode; I cancelled my plan. This time for good.

I'm ok if something is closed a reasonable amount but forcing people to nag update comments is just antagonistic to customers.

Yes, this exactly! It feels like they’re just creating more work for themselves.

From: Matt Rixman @.*>
Sent: Saturday, February 14, 2026 3:19 PM
To: anthropics/claude-code @.*>
Cc: cmwalton @.>; Comment @.>
Subject: Re: [anthropics/claude-code] [BUG] github-actions incorrectly closed a bunch of issues for "60 days of inactivity" (Issue #16497)

MatrixManAtYrService left a comment (anthropics/claude-code#16497) <https://github.com/anthropics/claude-code/issues/16497#issuecomment-3902541263>

Auto-closing issues causes duplicate issues to be created, so you then you lose the ability to count how many 👍's an issue has to determine how many people it affects. Instead you have to track down all of its duplicates and sum them.

Auto-locking issues prevents users from being able to leave comments that help each other. Unless some very specific kind of abuse is being prevented by locking issues, it should almost never be done.

I would suggest that Anthropic reconfigure these bots to apply/remove labels instead.

—
Reply to this email directly, view it on GitHub <https://github.com/anthropics/claude-code/issues/16497#issuecomment-3902541263> , or unsubscribe <https://github.com/notifications/unsubscribe-auth/BXGH35CTFSJVDJGEQSHLSH34L6GN7AVCNFSM6AAAAACQ3QVG22VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTSMBSGU2DCMRWGM> .
You are receiving this because you commented.Message ID: @.***>

https://github.com/anthropics/claude-code/issues/19267#issuecomment-3909971700 notes that this is now exacerbated by another bot which aggressively adds stale labels.

What I'd really like understand is, how has this issue not gotten any attention? Seems like there's been a lot of work towards finding ways to ignore the users. Has there been any corresponding work to identify high-profile problems and doing something about them?

For a company whose product is based on the idea that agents can handle it... Where is the agent for handling it? Like, I keep hearing about this culture where everybody is waking up to see all of this useful work that their agents have done while they sleep. But having something like this languish for over a month feels like a strong indicator that that work isn't actually as useful as the humans in charge think it is.

I take a more pessimistic view on the purpose of this repo. I don’t actually think they have anyone managing the issues and PR’s, AI or human, at all. This is just a repository to let us know when they’ve pushed a new release. So we’re diligently creating issues and PR’s and nothing ever comes of them. Just my two cents…

I don’t actually think they have anyone managing the issues and PR’s,

It's looking more and more like it, but as a matter of fact, I did receive attention from what appeared to be an employee to one of my tickets in the past.

Well, maybe I spoke too soon? Heh 🤣

This issue has been automatically locked since it was closed and has not had any activity for 7 days. If you're experiencing a similar issue, please file a new issue and reference this one if it's relevant.