Structuring LLM outputs: best practices for legal prompt engineering (2024) | Dark Hacker News