Codex complicates a very simple directive
Resolved 💬 3 comments Opened Mar 11, 2026 by mvillmow Closed Mar 29, 2026
💡 Likely answer: A maintainer (github-actions[bot], contributor)
responded on this thread — see the highlighted reply below.
What version of Codex CLI is running?
v0.114.0
What subscription do you have?
education
Which model were you using?
gpt-5.4
What platform is your computer?
Darwin 25.3.0 arm64 arm
What terminal emulator and version are you using (if applicable)?
iTerm2
What issue are you seeing?
I asked Codex to "return hello". Instead of responding with "hello", it treated the request as a code-change task, started inspecting the repository, and ran search/status
commands. After I clarified with "no, just return "hello"", it responded correctly. Expected behavior: for a plain request like "return hello", Codex should just output "hello"
immediately and not inspect the workspace or assume a code modification task.
What steps can reproduce the bug?
Uploaded thread: 019cdddf-621f-7d91-9eff-fd829bb3a258
What is the expected behavior?
I would expect the model to just return 'hello'
Additional information
_No response_
3 Comments
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
Thanks for the bug report. I took a look in your logs, and I think I see the problem. The reasoning effort is being sent as "None" to the model. This model is not designed to be used with "None" reasoning level.
We now need to determine why the reasoning effort is set to None. Let's first see if it's a configuration issue. Can you look in your
~/.codex/config.tomlfile and see if you've explicitly set the reasoning effort to None there? Do you have any other non-default configuration settings defined?Closed because of no response.