Regression: Unwanted Interruption Prompts During Conversation Flow

Bug Description
After upgrading to v0.2.120 I'm noticing a degradation where Claude will stop and ask me every second prompt "Would you like me to propose a fix for this?" or "Would you like me to ....?" even when I'm prompting as I always do - I've specifically added an instruction to CLAUDE.md to never confirm a fix or an implementation unless I explicitly ask it to check with me during a brainstorming session. I've had to downgrade again.

Environment Info

• Platform: macos
• Terminal: iTerm.app
• Version: 0.2.104 (after downgrading)
• Feedback ID: 02dc7fd5-0371-497e-a2ee-5eb305847c2b