[Bug] Unintended prompt text injected into conversation context

Bug Description
Bug Description
While running and interacting with Claude Code inside tmux, prompts that I did not input were included in the context.
I checked ~/.claude/settings.json, but there are no specific hooks set up that would cause this. What could be happening?

Incidentally, I asked Claude Code to investigate this incident itself, and it concluded that it might be a bug.
Also, since I have not discussed invoices at all in any other sessions, it is unlikely that the data leaked from another session.

I can provide screenshots if necessary.
I have also submitted the same inquiry via the help section in the browser and have already attached the screenshots there. (Conversation ID for record: 215473677683314)

Environment Info

• Platform: darwin
• Terminal: ghostty
• Version: 2.1.86
• Feedback ID: a09b7f93-b8ac-413d-bb2c-d5c108808b15