Fix App That Looks Good But Does Not Work

About this workflow

The systematic recovery flow for beautiful-but-broken AI builds: audit honestly, then fix by functional priority.

What happens when you run it

1. 1
   Get the honest map: what works, what is fake, what is broken.

   Runs flow-fix-app-looks-good-not-working-1-run-a-brutal-functional-audit on Claude Sonnet 4.6
2. 2
   Fix what matters most first - the app's core promise.

   Runs flow-fix-app-looks-good-not-working-2-prioritize-the-fix-order on Claude Sonnet 4.6
3. 3
   Make the single most important feature real, end to end.

   Runs flow-fix-app-looks-good-not-working-3-fix-the-core-value-action on Claude Sonnet 4.6
4. 4
   Convert every remaining FAKE and BROKEN item to WORKS.

   Runs flow-fix-app-looks-good-not-working-4-work-through-the-fix-queue on Claude Sonnet 4.6
5. 5
   Sweep the whole app for loading, empty, and error gaps.

   Runs flow-fix-app-looks-good-not-working-5-add-the-missing-states-layer on Claude Sonnet 4.6
6. 6
   Close the loop with the same brutal audit that started it.

   Runs flow-fix-app-looks-good-not-working-6-final-honest-re-audit on Claude Sonnet 4.6

Best for

• •Recurring tasks you currently copy-paste between tools.
• •When a single prompt isn't enough — you need a chain where output(n) feeds prompt(n+1).
• •Fork and adapt if your inputs differ materially from the author's use case.

Models used