[BUG] --worktree flag results in downgrade from --dangerously-skip-permissions in the middle of conversations

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?

When I invoke Claude Code with the --worktree flag and --dangerously-skip-permissions flag, at seemingly random points in the conversation, the --dangerously-skip-permissions is downgraded to accept edits on. This never happens when utilizing Claude Code without --worktree. I have also tried using without --worktree and asking Claude to create a worktree from within the session, and haven't seen this issue (but haven't done this extensively). When using Claude Code normally, --dangerously-skip-permissions is never downgraded.

What Should Happen?

--dangerously-skip-permissions should persist in sessions with the --worktree flag without any degradation

Error Messages/Logs

Steps to Reproduce

1. Invoke Claude Code via CLI with the claude --dangerously-skip-permissions --worktree flags.
2. Proceed with coding as normal.
3. Watch for degradation of --dangerously-skip-permissions to accept edits on, requiring user approval on every future step.

Claude Model

Opus

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

2.1.84

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

_No response_