This post outlines a simple view of prompt engineering for LLM CoT, arguing that concise instructions, divide-and-conquer task design, and better workflow decomposition can improve reasoning quality.