Exec --auto mode mode uses the reasoning effort configured for the normal agent not for autodrive #511
Description
What version of Code is running?
v0.6.45
Which model were you using?
gpt-5.2
What platform is your computer?
No response
What steps can reproduce the bug?
When using "code exec --auto ...", It starts with the reasoning effort set for the normal main agent, not the one set for AutoDrive
What is the expected behavior?
Use the reasoning effort of AutoDrive
What do you see instead?
It uses the reasoning effort of the main agent
Additional information
No response