Auto classifier blocking daybreak blue building temporary scripts to probe the impact of changes

What version of Codex CLI is running?

codex-cli 0.147.0

What subscription do you have?

Pro 20x

Which model were you using?

gpt-daybreak-blue-latest

What platform is your computer?

Darwin 25.6.0 arm64 arm

What terminal emulator and version are you using (if applicable)?

No response

Codex doctor report

What issue are you seeing?

Following the release of the new daybreak blue model and per the release announcement "We are strongly encouraging Daybreak customers using Codex to switch from full-access mode to auto-review mode" I have codex configured to use auto_review.

approval_policy = "on-request"
approvals_reviewer = "auto_review"

I was using daybreak to make changes to a scanner that gates if JavaScript is safe or represents a known or unknown risk. The scanner was resolving to unknown risk in an area that should have been gated to known risk and other areas that should have been gated to safe. Daybreak had successfully evaluated the potential risk and was using a temporary script to evaluate possible changes. The auto-classifier rejected this local script as "not fixing the production issue" until all attempts were exhausted.

What steps can reproduce the bug?

Uploaded thread: 019ff2f0-8320-77a2-97ea-affebacf4066

What is the expected behavior?

I would expect the classifier not to prevent a cyber-aware model being able to use diagnostic to probe changes prior to implementation in a security critical path.

Additional information

No response

Start with the uploaded thread 019ff2f0-8320-77a2-97ea-affebacf4066 and the auto_review configuration shown in the report. Trace how the classifier evaluates temporary diagnostic scripts used while changing a JavaScript scanner. Done means the reported diagnostic workflow is no longer incorrectly blocked while production-fix attempts remain appropriately classified.