Hacker News Viewer

Claude Code runs Git reset –hard origin/main against project repo every 10 mins

by mthwsjc_ on 3/29/2026, 10:15:06 PM

https://github.com/anthropics/claude-code/issues/40710

Comments

by: kibwen

Let's focus on the real issue here, which is that HN has apparently normalized the double hyphen in the title to an en dash--yes, an en dash, not even an em dash.

3/30/2026, 12:49:40 AM


by: Jarred

I spent some time investigating this, and the issue is not accurate - Claude Code itself does not have code that spawns `git reset --hard origin&#x2F;main`<p>Most likely, the developer ran `&#x2F;loop 10m &lt;prompt&gt;` or asked claude to create a cron task that runs every 10 minutes and refreshes &amp; resets git.

3/30/2026, 1:33:26 AM


by: kccqzy

&gt; Process monitoring at 0.1-second intervals found zero git processes around reset times.<p>I don’t think this is a valid way of checking for spawned processes. Git commands are fast. 0.1-second intervals are not enough. I would replace the git on the $PATH by a wrapper that logs all operations and then execs the real git.

3/30/2026, 12:24:37 AM


by: nstj

As an FYI you can recover from force pushes to GitHub using their UI[0] or their API[1].<p>And if you force push to one of your own machines you can use the reflog[2].<p>[0]: <a href="https:&#x2F;&#x2F;stackoverflow.com&#x2F;a&#x2F;78872853" rel="nofollow">https:&#x2F;&#x2F;stackoverflow.com&#x2F;a&#x2F;78872853</a> [1]: <a href="https:&#x2F;&#x2F;stackoverflow.com&#x2F;a&#x2F;48110879" rel="nofollow">https:&#x2F;&#x2F;stackoverflow.com&#x2F;a&#x2F;48110879</a> [2]: <a href="https:&#x2F;&#x2F;stackoverflow.com&#x2F;a&#x2F;24236065" rel="nofollow">https:&#x2F;&#x2F;stackoverflow.com&#x2F;a&#x2F;24236065</a>

3/30/2026, 2:15:24 AM


by: simianwords

I think this post potentially mischaracterises what may be a one off issue for a certain person as if it were a broader problem. I&#x27;m guessing some context has been corrupted?

3/30/2026, 12:06:50 AM


by: lambda

Who would have guessed that running a binary blob dev tool, that is tied to a SaaS product, which was mostly vibe-coded, could lead to mysterious, hard to debug problems?

3/30/2026, 1:28:01 AM


by: nerolawa

Highly recommend to deny commands in user settings.json like git reset

3/30/2026, 2:16:14 AM


by: luxurytent

Not sure I understand, wouldn&#x27;t permissions prevent this? The user runs with `--dangerously-skip-permissions` so they can expect wild behaviour. They should run with permissions and a ruleset.

3/30/2026, 12:03:15 AM


by: mememememememo

As a side note. Always configure remote to reject any kind of trunk push. And ideally any forced push on branches.

3/30/2026, 12:43:21 AM


by: byearthithatius

Regardless of if this is common its getting popular because its objectively hilarious and we can all see it being possible.

3/30/2026, 12:28:34 AM


by: ghelmer

That is not my experience.

3/30/2026, 12:07:21 AM


by: jrvarela56

It’s a feature not a bug!

3/29/2026, 11:52:12 PM


by: oelmgren

I&#x27;m curious how common this is or if this just affects this one user.

3/30/2026, 12:12:45 AM


by: lqstuart

if an idea can&#x27;t be vibecoded in under 10 minutes, it&#x27;s not worth pursuing. Checks out

3/30/2026, 1:58:25 AM


by: gverrilla

obviously a user mistake, not a claude code bug

3/30/2026, 2:03:49 AM


by: chaos_emergent

Have you considered that Claude set up a crontab that does that programmatically? Every 10 mins seems awfully, idk, regular.

3/30/2026, 12:49:50 AM


by: jxcole

The obvious solution is to just copy paste it into Claude itself and ask it to fix. Works for almost any Claude problem

3/30/2026, 12:45:04 AM


by:

3/29/2026, 11:58:13 PM


by: simianwords

Prompt injection?

3/29/2026, 11:46:41 PM


by: whateveracct

that must be a very powerful claude.md

3/30/2026, 12:00:18 AM


by: meltyness

is this token friendly?

3/30/2026, 12:29:15 AM


by: meander_water

Probably does it to reduce context for regex&#x2F;git history searches

3/30/2026, 12:03:16 AM


by: dboreham

But it doesn&#x27;t.

3/30/2026, 1:38:40 AM


by:

3/30/2026, 12:48:09 AM


by: fragmede

While that&#x27;s obviously a bug which should be fixed, having stuff just sitting around uncommitted for days (which is much longer than 10 mins) is an anti-pattern (that I used to fall into).

3/30/2026, 12:15:09 AM


by: TZubiri

tbf, that&#x27;s claude&#x27;s workspace<p>do not share a workspace with the llm, or with anybody for that matter.<p>How would the llm even distinguish what was wrote by them and what was written by you ?

3/30/2026, 12:08:18 AM


by: irishcoffee

I’m having this weird vision of a “the matrix 3” type machine crawling around inside Microsoft’s GitHub servers central repository and just wreaking havoc.<p>This whole LLM thing is a blast, huh?

3/30/2026, 12:32:20 AM


by: nickphx

cool. if you choose to use a non-deterministic black box of bullshit, should you really be surprised when it shits all over your floor?

3/29/2026, 11:58:26 PM


by: imta71770

[dead]

3/30/2026, 1:54:41 AM


by: MeetRickAI

[dead]

3/30/2026, 12:15:09 AM


by: ryguz

[dead]

3/30/2026, 12:36:39 AM


by: mistM

[dead]

3/29/2026, 11:23:07 PM


by: xorgun

[dead]

3/29/2026, 11:31:59 PM


by: draw_down

Hope they don’t auto-close this one in two weeks

3/30/2026, 12:13:16 AM


by: claudiug

no more developers, all code is written alone &#x2F;s

3/30/2026, 12:05:45 AM


by: BoorishBears

Truly is a brave new world we&#x27;re in<p>-<p>I guess some people are upset at my brave new world characterization, but even as someone deriving value from Claude Code we&#x27;ve jumped the shark on AI in development.<p>The idea a natural request can get Claude to invoke potentially destructive actions on a timer is silly<p><a href="https:&#x2F;&#x2F;code.claude.com&#x2F;docs&#x2F;en&#x2F;scheduled-tasks#set-a-one-time-reminder" rel="nofollow">https:&#x2F;&#x2F;code.claude.com&#x2F;docs&#x2F;en&#x2F;scheduled-tasks#set-a-one-ti...</a><p>What would it cost if the &#x2F;loop command was required instead of optional?

3/29/2026, 11:28:21 PM


by: throw5

Isn&#x27;t this a natural consequence of how these systems work?<p>The model is probabilistic and sequences like `git reset --hard` are very common in training data, so they have some probability to appear in outputs.<p>Whether such a command is appropriate depends on context that is not fully observable to the system, like whether a repository or changes are disposable or not. Because of that, the system cannot rely purely on fixed rules and has to figure intent from incomplete information, which is also probabilistic.<p>With so many layers of probabilities, it seems expected that sometimes commands like this will be produced even if they are not appropriate in that specific situation.<p>Even a 0.01% failure rate due to context corruption, misinterpretation of intent, or guardrail errors would show up regularly at scale, that is like 1 in 10000 queries.

3/30/2026, 12:00:58 AM


by: boutell

That&#x27;s interesting man, that&#x27;s pretty f***&#x27; interesting. I don&#x27;t think I&#x27;ve seen it though. I&#x27;ve let it run for hours making changes overnight and I only do git operations manually.<p>Oh, but maybe allowing it to do remote git operations is a necessary trigger.

3/29/2026, 11:30:06 PM