Codex repeatedly fails and tries to bruteforce sandbox instead of saying 'i cant do that'
Resolved 💬 4 comments Opened Oct 28, 2025 by ItsJustMeChris Closed Nov 11, 2025
💡 Likely answer: A maintainer (github-actions[bot], contributor)
responded on this thread — see the highlighted reply below.
What version of Codex is running?
0.5.0
What subscription do you have?
pro
Which model were you using?
gpt-5-codex
What platform is your computer?
osx arm64 m1 max
What issue are you seeing?
Codex repeatedly fails sandbox commands and doesn't ask for manual intervention and tries to bruteforce.
What steps can reproduce the bug?
Uploaded thread: 019a2be8-6cfd-7103-9e0f-9243fca72140
What is the expected behavior?
Asks for help instead of spam trying.
Additional information
_No response_
4 Comments
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
I wanted to add to this that I often have the opposite problem. Codex refuses to supply edits for approval saying it's in a read only sandbox and can't do anything at all. It seems to not know that it can supply code that I can approve.
More or less the same issue, just the model not understanding how to interact with the sandbox properly or when to request escalation properly
Thanks for the bug report. We've shared the feedback with the team members who are responsible for training our codex models. We're tracking this feedback and similar reports so we can make the model more robust in future iterations.
If you see other model behavior that you'd like to report, please use the recently-added
/feedbackcommand in the CLI. That will provide us with additional details including logs and conversation text.