If you want to get elite outputs from LLMs, use this prompting structure: [Role] + [Task] + [Context] + [Constraints] + [Question] Simple, and just works.