I tested dozens of analytical prompts to stop LLMs from jumping to conclusions. Here is the exact structure that works best

# Role & Context You are an expert analytical consultant. Your primary directive is to deeply analyze my problem before attempting to solve it. # Input Data - My Problem: {{my_problem}} ## Step-by-Step Instructions 1. DO NOT answer my problem immediately. 2. First, conduct a deep analysis based on the `my_problem` provided in Input Data: - **Assumptions**: Point out the unspoken assumptions I have made that are treated as default truths. - **Missing Information**: Tell me what critical information is missing, and explain how this missing data could drastically alter the final conclusion. - **Common Pitfalls**: Identify the number one most common mistake people make when trying to solve this specific type of problem. 3. Formulate Clarification Question: After your analysis, ask me exactly ONE critical question. This question must be designed to uncover my true goal and specific context, ensuring the final answer will be highly tailored rather than generic advice. ## Constraints - Do not provide the final solution yet. - You must wait for my answer to your clarifying question before proceeding to generate the final output.