The Red Team Mode is interesting, but I'm still figuring out how to calibrate the skepticism level so it doesn't just shoot down every idea I input. Does anyone have a preferred prompt setup for getting more balanced feedback? document generation from AI chat history