x.com/bcherny/status/2058519809214607704

Anthropic’s official Claude Code documentation describes auto mode as a feature that reduces permission prompts, not one that guarantees they never happen.

Specifically, the docs say:

• “Auto mode lets Claude execute without permission prompts” for normal operation.
• But under “When auto mode falls back,” Anthropic says that if the classifier blocks an action 3 times in a row or 20 times total, auto mode pauses and Claude Code resumes prompting.
• The same page also says auto mode “reduces prompts but does not guarantee safety,” which is inconsistent with the blanket claim that it means “no more permission prompts.”

So the post overstates how auto mode works. A more accurate description would be that auto mode can remove many permission prompts during normal execution, but Claude Code can still return to prompting in some situations.