Prompt upgrader — diagnose why it's weak, then fix it with a scored A/B comparison

Already have a prompt that's not quite working? Paste it and get a full diagnosis + upgrade:

'You are an expert prompt engineer who debugs prompts the way a developer debugs code — by identifying the specific failure mode first, then fixing it.

My current prompt: [PASTE YOUR PROMPT].
What I wish it did better: [e.g. too generic / too long / wrong format / inconsistent results].
I'm using it on: [ChatGPT / Claude / Gemini / Midjourney / etc.].

STEP 1 — DIAGNOSE. Score my prompt on each dimension (1-5) and identify the specific failure mode(s):
• Role clarity (1 = no role or generic, 5 = specific expertise + thinking style defined)
• Task specificity (1 = vague ask, 5 = unambiguous deliverable)
• Constraints & guardrails (1 = none, 5 = prevents rambling, wrong format, and off-topic drift)
• Output format (1 = unspecified, 5 = structured, immediately actionable)
• Examples / few-shot (1 = none when needed, 5 = well-chosen examples that anchor quality)
• Model fit (1 = prompt fights the model's strengths, 5 = plays to them)

STEP 2 — REWRITE. Produce the upgraded prompt with [BRACKETS] for customizable parts. The rewrite must fix every diagnosed weakness.

STEP 3 — A/B COMPARISON. Show the original and upgraded prompts side by side in a table:
| Dimension | Original (score) | Upgraded (score) | What changed & why |
Annotate every meaningful change.

STEP 4 — ONE MORE LEVEL. Suggest one advanced technique that could push this prompt further (chain-of-thought, self-critique loop, multi-pass, few-shot examples, temperature guidance, or a verification step).'

Tips: include a sample of the BAD output the current prompt produces — this gives the AI a concrete target to fix; specify which model you're using because prompt strategies differ (Claude rewards 'think step by step,' Gemini leverages large context, ChatGPT responds well to persona); if the prompt is for image generation, mention the platform so the rewrite uses correct parameters.