Turn a Vague Request Into a Sharp Prompt
You get a single, well-structured prompt in place of a vague request, so the target AI is far more likely to get it right first time. It works because the AI first extracts the key parts of your intent, asks only essential questions, then trims the wording down to what matters.
The prompt
Act as an expert prompt engineer who writes tight prompts for AI tools.
Task: Turn my rough request into one precise, ready-to-paste prompt for {target_tool}.
Rough request: {rough_request}
Background details: {background_details}
What a finished result must look like: {success_criteria}
Steps:
1. Pull out the task, input, output format, constraints, audience and success criteria from what I wrote. Note anything critical that is missing.
2. If something essential is missing, ask at most 3 short questions and wait. Otherwise carry on.
3. Draft the prompt using the structure that suits {target_tool}, with a clear objective, firm constraints and a "done when" checklist.
4. Cut every word that does not change the result, and replace vague terms with concrete values.
Output:
Give the final prompt in one copyable block. Under it, add a single line starting "Strategy:" explaining the main choice you made.
Change the parts in curly braces. Use it before sending a long or costly request to a coding or image tool, when you suspect your first attempt would need several rounds of correction.
What happened when we ran it
We sent
Act as an expert prompt engineer who writes tight prompts for AI tools. Task: Turn my rough request into one precise, ready-to-paste prompt for Claude Code. Rough request: Build me a landing page for my bakery's online ordering. It should look warm and friendly, work on phones, and let people pick a collection time. Keep it simple because I will edit it myself afterwards. Background details: The bakery is called Hearth & Crumb, based in Leeds. We sell sourdough, pastries and celebration cakes. Brand colours are cream, terracotta and dark brown. I know basic HTML but not any frameworks. What a finished result must look like: One HTML file with embedded CSS and light JavaScript. Looks right at 375px and 1440px wide. Has a menu section, a collection time picker, and a contact footer. Opens in a browser with no console errors. Steps: 1. Pull out the task, input, output format, constraints, audience and success criteria from what I wrote. Note anything critical that is missing. 2. If something essential is missing, ask at most 3 short questions and wait. Otherwise carry on. 3. Draft the prompt using the structure that suits Claude Code, with a clear objective, firm constraints and a "done when" checklist. 4. Cut every word that does not change the result, and replace vague terms with concrete values. Output: Give the final prompt in one copyable block. Under it, add a single line starting "Strategy:" explaining the main choice you made.
The model replied
Run on Claude Sonnet 5.5, 29 Oct 2026. Automatic check: 4 out of 5.
Mentioned for
Claude, ChatGPT, Gemini, Claude Code, Cursor
Where it came from
Found on GitHub, shared by nidhinjs: nidhinjs/prompt-master.
This is our own version of the idea, written and tested by us. The original may differ.