Prompt Debugging Interview
Find out why a prompt gives poor results.
The Open buttons send the prompt text as currently shown below to that site only when you click. Fill in the placeholders first.
Help me debug a prompt by interviewing me first. Collect the inputs by asking me one thing per turn: (1) the prompt exactly as I used it, (2) the output I got, (3) the output I wanted (describe or give an example), (4) how often it fails (every time or sometimes), (5) what I changed recently, (6) the tool and settings used, and (7) whether any input is long, messy, or sensitive. What I have: [PROMPT_AND_OUTPUT]. After the interview, diagnose using this checklist and report each as likely or unlikely with evidence from my prompt: ambiguity in the task; missing context or audience; conflicting instructions; format not specified or too complicated; too many tasks in one prompt; missing examples where pattern matters; input too long or messy; instructions buried in the middle; roles or tone distracting from the task; requests for information the model cannot know; and settings differences. Then return a revised prompt, a list of changes with reasons, three test inputs to compare old and new, a way to judge the difference with a rubric, and the next experiment if it still fails. Do not rewrite the prompt before you have asked the questions; do not promise that a prompt will work every time.
Fill in before use
Optional: type your details and the prompt above updates. Nothing is saved or sent.
Replace each bracketed placeholder with your own details, then paste into Any AI or another assistant.