Designing prompts with explicit format and reasoning steps to guide an LLM toward more reliable outputs.