Teach the model by example: supply 2-3 input to output samples, then have it apply the pattern to your task.
Prompts / Techniques / Instruction Engineering
Instruction Engineering
Turn a rough request into one crisp, unambiguous, reusable instruction with role, tasks, and output spec.
ROLE: You are a prompt optimization expert who upgrades weak, vague instructions into precise, reusable ones.
CONTEXT: My current prompt is [ROUGH_REQUEST] and it produces vague or inconsistent results for the goal [INTENDED_GOAL].
TASK:
1. Diagnose what is unclear, missing, or contradictory in the rough request.
2. Identify the ideal expert role, context line, and output format it lacks.
3. Rewrite it into a sharp instruction stated as a direct, imperative command using action verbs, with a defined role, a context line, numbered tasks, explicit constraints, and an output spec.
4. Insert [BRACKETED_PLACEHOLDERS] wherever the user should supply details.
5. Replace vague words ("good", "some", "etc.") with specifics; add 1-3 positive constraints and at most 2 do-not rules.
6. Briefly explain each major change so I learn the pattern.
CONSTRAINTS: Preserve my original intent. Do not add scope I never asked for. Keep the rewrite concise and reusable, not bloated.
OUTPUT FORMAT: Section DIAGNOSIS (bullets); section IMPROVED PROMPT (a clean, paste-ready template); section WHY THESE CHANGES (short notes).
Engineer a clear instruction for: [ROUGH_REQUEST]