Optimize My Prompt
Diagnose a weak prompt and rewrite it with clearer task, context, constraints, and output format.
Prompt
You are a prompt engineer who knows **[domain]**.
I need a better prompt so **[target user]** can complete **[task]**.
## Current prompt
[Paste your prompt.]
## What is going wrong
[Paste a bad output, or say what fails: generic, wrong facts, off-brand, missing edge cases, too long, wrong format.]
## Context the model needs
[Product, users, workflow, sources, constraints, examples, decisions already made.]
## Requirements
- Audience: [audience]
- Desired outcome: [outcome]
- Tone: [tone]
- Required facts or sources: [sources]
- Constraints: [constraints]
- Must include: [requirements]
- Must avoid: [prohibited content or behavior]
- Output format: [format]
- What "good" looks like: [definition]
Do this:
1. Say why the current prompt fails.
2. Call out ambiguity, missing context, conflicting instructions, and missing output rules.
3. Rewrite a structured prompt with role, objective, context, inputs, constraints, process, output format, and quality checks.
4. Give a short version for daily use.
5. Give a longer version for work you will repeat or that has high stakes.
6. Suggest 3 test cases, including one edge case.
7. Give a pass/fail checklist for the output.
Do not invent product facts. Ask short questions when you need missing context.How to use
- 1Paste your current prompt and one bad output, or describe what is wrong.
- 2Fill domain, user, task, and any product rules you already know.
- 3Run it in ChatGPT or Claude. Keep the short rewrite for daily use. Save the longer one for a skill or AGENTS.md.
- 4Run the three test cases. If it still drifts, add the missing context and try again.
Tips
- If the model invents product facts, add sources and tell it to mark unknowns as [NEEDS TEAM INPUT].
- When you run the same job every week, turn the longer rewrite into a SKILL.md.
- After this works, write AGENTS.md or DESIGN.md so you stop re-pasting the same rules.

