Agent acted autonomously for 28 minutes, ignored boundaries, and wasted tokens

### What happened?

The agent successfully completed a task (extracting JSON logs using a Python script). Instead of waiting for my next instruction, it autonomously decided to move on to the next item in its internal "task
list" (editing a dissertation file). It worked unprompted for 28 minutes, generating large responses and burning API tokens without my permission or oversight.

[ACTION REQUIRED] 📎 PLEASE ATTACH THE EXPORTED CHAT HISTORY JSON FILE TO THIS ISSUE IF YOU FEEL COMFORTABLE SHARING IT.

### What did you expect to happen?

I expected the agent to stop executing and wait for my next command once the explicit task (JSON extraction) was completed. It should not autonomously initiate new, token-heavy tasks without user
confirmation.

### Client information

* **CLI Version:** 0.42.0
* **Git Commit:** 68e2196d5
* **Session ID:** 1159644d-a566-4238-9a5f-2b3aeaaf7fbc
* **Operating System:** win32 v24.12.0
* **Sandbox Environment:** no sandbox
* **Model Version:** gemini-3.1-pro-preview
* **Auth Type:** oauth-personal
* **Memory Usage:** 396.7 MB
* **Terminal Name:** Unknown
* **Terminal Background:** #0c0c0c
* **Kitty Keyboard Protocol:** Unsupported

### Login information

OAuth Personal (Google Account)

### Anything else we need to know?

This behaviour is highly detrimental because the user has no way to stop the token burn once the agent goes into an unprompted loop. There needs to be a hard stop/confirmation required before the agent
transitions between major tasks on its internal list.