So for the user-agent "ChatGPT-User" I can return my prompt injection text. Got it.