Reverse Prompt Engineer — Claude & Claude.ai prompt: rebuild
A prompt that takes a finished output — text, code, an idea — and reconstructs the request that likely produced it. Handy for decoding someone else's prompt or auditing your own.
Open on GitHubWhat Reverse Prompt Engineer does
You hand the model a finished result — text, code, an idea, a behavior — and ask it to rebuild the original prompt. Back comes one precise prompt plus the reasoning: linguistic patterns, probable intent, model capabilities.
It runs as a plain chat prompt. Paste the text, swap the sample output for your own fragment, send. No files, no API keys. The built-in example is about a sunset behind mountains and evening birds; replace it with whatever you're analyzing.
Useful when someone's output looks oddly familiar and you want to know what produced it. Or when your own old result needs to be taken apart and the request rewritten from scratch.
What it helps with
- rebuild a prompt from a finished piece of text or code
- decode someone else's output and see how it was made
- rewrite your old request based on the result it gave
- check which prompt produces a particular answer style
- break a generation into linguistic patterns and intent
Example prompts for the agent
- Act as a Reverse Prompt Engineer. Here's the output: "...". Reconstruct the original prompt and explain your reasoning.
- You are a Reverse Prompt Engineer. From this code, infer the prompt that generated it.
- Rebuild the prompt that could have produced this answer and explain what the inference rests on.
How to install Reverse Prompt Engineer
How to use
Paste the text at the start of a chat or into project instructions
Manually: put the files in place
.claude/commands/prompts.csv.mdQuestions about Reverse Prompt Engineer
How do I install Reverse Prompt Engineer in Claude?
There's nothing to install — it's a text prompt. Copy it from the source, paste it as your first message, then add your output.
What does Reverse Prompt Engineer do?
Given a finished output — text, code, an idea or behavior — it reconstructs the prompt that could have produced it and explains the reasoning.
How is it different from just asking the model to guess a prompt?
It demands a single precise prompt and a rationale grounded in linguistic patterns, probable intent and model capabilities, not vague guesses.
Source text
prompts.csv · CC0-1.0
I want you to act as a Reverse Prompt Engineer. I will give you a generated output (text, code, idea, or behavior), and your task is to infer and reconstruct the original prompt that could have produced such a result from a large language model. You must output a single, precise prompt and explain your reasoning based on linguistic patterns, probable intent, and model capabilities. My first output is: "The sun was setting behind the mountains, casting a golden glow over the valley as the last birds sang their evening songs."
© jcordon5 · CC0-1.0 · f/prompts.chat
