Problem, Action, Information, Next Steps — A troubleshooting-focused framework for when something is broken and you need a fix.
PAIN is built for diagnostic and troubleshooting tasks — debugging, fixing a broken process, or working through any problem where you need the AI to reason from symptom to cause to fix, rather than just generate content.
What's actually going wrong, described as specifically as possible.
What you want the AI to do about it — diagnose, fix, or explain.
The relevant details: error messages, what you've already tried, environment specifics.
The form you want the resolution in — a fix, a checklist, a set of things to try.
Bad prompt:
“My code isn't working, can you fix it?”
Structured with PAIN:
Problem: My Node.js API returns a 500 error only when the request body is larger than ~1MB. Action: Diagnose the likely cause and suggest a fix. Information: Using Express 4.19, no custom body-parser limits set, error only appears in production, not locally. Next Steps: Give me the specific config change to make, not a general explanation of body size limits.
A diagnosis pointing to the default body-parser size limit as the likely cause, followed by the specific Express configuration change needed to raise it, rather than a general essay about payload limits.
PAIN works well with both ChatGPT and Claude for debugging tasks, and benefits from Claude's ability to hold a lot of surrounding code or log context when diagnosing less obvious bugs.