The ultimate LLM prompt: extract the best possible answers with the highest fidelity and lowest error rates