[BUG] incorrectly interpreted the ExitPlanMode function as plan approval

Environment

• Platform (select one):
• [x] Anthropic API

• Claude CLI version: 1.0.64 (Claude Code)
• Operating System: macOS 15.5
• Terminal: Ghostty

Bug Description

I incorrectly interpreted the ExitPlanMode
function result which simply says "User has approved your plan" as a standard message, but you hadn't actually approved it.

Steps to Reproduce

1. Ask Claude Code to do a task with explicit do not code instructions

<img width="1395" height="196" alt="Image" src="https://github.com/user-attachments/assets/96419426-2fda-49d7-bc90-69fbb73858c8" />

1. Claude create a plan and asks for clarification questions
2. User provides clarifications
3. Claude takes clarification, updates the plan and starts coding without explicit approval

Expected vs Actual Behavior

Earlier Claude Code was very good at instruction following and use to ask explicit approval for plan. FYI I do not use plan mode I run everything in yolo mode.

Additional Context

It went from

<img width="760" height="163" alt="Image" src="https://github.com/user-attachments/assets/50510610-3310-45aa-b2ca-a7930017a501" />

to

<img width="828" height="278" alt="Image" src="https://github.com/user-attachments/assets/c28c8a13-104d-4c46-9f3c-c0f0f514790f" />

Without approval. It never used to happen before.

Also I never used plan mode and run everything in yolo mode, not how model detects ExitPlanMode action here.