Doom loop requesting permission for trivial file edits
Open 💬 2 comments Opened Dec 27, 2025 by ianibo
What version of Codex is running?
>_ OpenAI Codex (v0.77.0)
What subscription do you have?
Plus with additional usage credits
Which model were you using?
GPT5.2
What platform is your computer?
Linux 6.17.0-8-generic x86_64 unknown
What issue are you seeing?
Repeated prompts for permission to edit a file despite several requests to stop asking
What steps can reproduce the bug?
Uploaded thread: 019b6143-8191-7a83-9143-086229565ca7
What is the expected behavior?
Get on with trivial edits
Additional information
_No response_
2 Comments
Hello everyone, can you please check this https://github.com/openai/codex/pull/11868
I've reproduced the issue, added a test, and verified that the test fails before my changes and passes after them.
This feels like two guards fighting each other: loop protection and approval memory.
For trivial edits, I’d cache a scoped approval lease (
same file,same mutation class, short TTL) so the agent does not keep re-asking while still preserving a hard boundary when scope widens. Repeated prompts like this are often an approval-state bug, not a real new safety event.