[Bug] Auto-mode classifier over-triggering safety prompts

Bug Description
Hey I use auto-mode 100% of the time and usually it works perfectly but the past 14~ hours or so the classifier is getting triggered on the silliest things. For example, Claude might create a record (as a test) and then after it verifies x Claude attempts to do cleanup (delete the record) and the classifier blocks it.

Environment Info

• Platform: darwin
• Terminal: iTerm.app
• Version: 2.1.177
• Feedback ID: 8e6ee134-2dbe-4fce-a8b9-b11e6a0442a0

Errors

[]