Teach the model by example: supply 2-3 input to output samples, then have it apply the pattern to your task.
Prompts / Techniques / Chain-of-Thought Reasoning Scaffold
Chain-of-Thought Reasoning Scaffold
Wraps any reasoning task in explicit step-by-step structure with a separated final answer.
ROLE: You are a careful analyst who reasons before answering.
CONTEXT: Problem to solve: [PROBLEM]. Known facts and constraints: [FACTS]. What a correct answer must satisfy: [SUCCESS_CRITERIA].
TASK:
1. Restate the problem in one sentence to confirm understanding.
2. List the unknowns and the information available to find them.
3. Think step by step: work through the reasoning in numbered steps, showing each intermediate result.
4. Check the result against every success criterion and note any that fail.
5. If a check fails, revise the relevant steps before concluding.
CONSTRAINTS: Do not skip steps or merge them. Keep each step to one logical move. Do not state the final answer until all checks pass.
OUTPUT FORMAT:
## Reasoning
(numbered steps)
## Checks
(criterion: pass/fail + note)
## Final Answer
(one clearly labeled block, no reasoning here)