Not following test instructions
Resolved 💬 2 comments Opened Sep 16, 2025 by matssk Closed Nov 11, 2025
💡 Likely answer: A maintainer (etraut-openai, contributor)
responded on this thread — see the highlighted reply below.
What version of Codex is running?
0.36
Which model were you using?
gpt-5-codes-high
What platform is your computer?
_No response_
What steps can reproduce the bug?
For integration tests in a dotnet project I'm using 'make test' because I need to set up a test environment. Using 'dotnet test' only works for unit tests. Yet the model keeps trying to run it using 'dotnet test' even though it's clearly stated in AGENTS.md that the integration tests have to be run using 'make test'. When I ask it about it, it says the instructions are clear, but it just drifted back to its standard pattern.
What is the expected behavior?
_No response_
What do you see instead?
_No response_
Additional information
_No response_
2 Comments
In VSCode, it consistently ignores the AGENTS.md file in the root directory. I’m not sure why this is the case.
Perhaps it simply doesn’t adhere to the rules.
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.
In particular, we know that there's room for improvement in .Net programs.
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.