I've been able to get LLMs to push back on ideas by just adding language to the system prompt requiring that they adopt a skeptical persona (insert whatever per

I've been able to get LLMs to push back on ideas by just adding language to the system prompt requiring that they adopt a skeptical persona (insert whatever persona you want for your use case). I see the word "skeptical" appear in their thought processes as a result, and my anecdotal experience is that they are less agreeable as a result. People need to put more thought into what these systems are and what they can do to help shape their output.