What this is for
Walkthroughs end with twenty photos and zero structure. By the time you’re back at the truck, half the items are forgotten, the other half are described differently than what you saw. This prompt forces structure on the photo dump and — more importantly — refuses to invent items it can’t see.
How to use it
- At the end of the walkthrough, take wide-shot, mid-shot, and close-up photos of every issue. Number them in order.
- Upload the photos to ChatGPT (vision), Claude (vision), or Gemini.
- Paste the prompt. Fill in the project context line.
- Review the output and cross-check against your memory of the walkthrough. The “needs verification” items are usually where the value is.
Pro move
Add a final line to the prompt: “For any P1 item, draft a 2-sentence text I can send the GC right now.” Now you walk back to the truck with the punch list AND the texts already drafted. Send them before you drive off the site.