[BUG] Safeguard triggering over literally nothing

Preflight Checklist

• [x] I have searched existing issues and this hasn't been reported yet
• [x] This is a single bug report (please file separate reports for different bugs)
• [x] I am using the latest version of Claude Code

What's Wrong?

Fable was working in automatic mode to create setting to lock mobile app in portrait mode. This is as boring and mundane as it can be. No conversation with user, the app itself is boring technical tool. I asked Opus what was the last message before it happened, here is the answer:

"Precisely: the switch notification was delivered attached to the tool result of my PowerShell call that opened the Appearance page and scrolled down to the Orientation section — the command that returned saved and was about to reveal the "Lock to portrait" toggle. It came in through the "The user sent a new message while you were working" channel on that tool result.

So it arrived against my tool activity, not after a fresh user message. The immediately preceding actions in the transcript were: pushing a lockPortrait:true settings.json into the app via run-as, launching the app, taking screencaps to confirm the app stayed portrait, then tapping into Settings → Appearance and scrolling to the Orientation section."

---

Result: I'm switching to OpenAI immediately when Fable becomes separately paid. I am fed up with this shit.

What Should Happen?

Claude should have been working.

Error Messages/Logs

It doesn't create any logs.

Steps to Reproduce

Impossible to tell, mechanism is deliberately hidden by Anthropic.

Claude Model

Other

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

latest

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

PowerShell

Additional Information

_No response_