Codex is reverting unrelated changes made by other AI agents
I'm running into a very frustrating issue with Codex. It keeps telling me, "I noticed there are files I didn't touch. Do you want me to do something about them?" I say no, but then I also find it undoing changes that other AI agents made. For example, if I'm running multiple Codex instances in parallel, or maybe Cursor agent, it sees the changes in the repo unrelated to its work and reverts them without telling me -- even after I told it: "Do NOT change any files you did not touch. NEVER do that!!!"
This is absolutely infuriating. It makes Codex much less useful and much less trustworthy. I've updated my AGENTS.md file to explicitly say not to do this, but it still happens. It seems like this is some sort of system instruction that you guys have put in there, and it's idiotic.
PLEASE FIX THIS. I AM LOSING MY MIND OVER HERE.
5 Comments
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
you opened 10 same issues?
have you heard of https://git-scm.com/docs/git-worktree
Maybe opening a few more cases next time will help?
What you're seeing here is model behavior, not something that can be changed with a simple bug fix in the Codex harness. We'll share the feedback with the team responsible for training our codex models. If you see other model behavior that you'd like to report, please use the
/feedbackcommand. That will provide us with additional details including logs and conversation text.@etraut-openai thx for the quick response. i did that. After submitting the feedback, it then pointed me here.
Sorry again about all the issue duplication. I genuinely do not know how that happened. 🤷♂️