ChatGPT's first answer is usually agreement. Ask it whether your plan is good and it will find reasons the plan is good. That is not analysis, it is a mirror.
This prompt splits the response in two. It has to attack the plan before it is allowed to give you an opinion, so you see the case against before you see the recommendation.
The prompt
Before you answer, argue against my plan as convincingly as you can, like you're trying to talk me out of it. List every risk and weak assumption. Only after that, tell me what you'd actually do and why.Paste it above your plan, or send it first and paste the plan in your next message.
How to use it
Works on decisions, not facts. Use it for a strategy, a hire, a launch date, a pricing change.
Give it something to argue with. A one-line plan gets a one-line objection.
Read the objections before the recommendation. If you skip to the bottom you have wasted the prompt.
What it doesn't do
It doesn't make the model right. You get a well-argued case against your plan whether or not the plan is bad, so treat the objections as a checklist to test, not a verdict to accept.