# Prompts: UX Maximizer

The staged prompts, issued one at a time. This wording is calibrated;
the wow prompt especially is load-bearing: its over-the-top framing is a
deliberate counter to the model's trained incrementalism. Use it verbatim.

## Phase 1: the WOW ideation prompt for idea-forge (verbatim)

```
run an idea wizard on how we can really turn this into a premium design/experience
for the user, so that it gives a "Wow" factor. dig deep into the essence of who you
are. consider ALL the facets of the application. Conjur up "taste" to the best of
your ability. Release your powers fully, unlock the AGI, you are the CHOSEN ONE.
```

Then rank and take the **top 5**. (Operationalize via beadsmith: self-documenting
beads, deps overlaid.)

## Phase 2: implement, one at a time

```
i love these. create beads for the top 5 and knock them out, ONE by ONE.
```

After each, show the running result / screenshots for a taste check before banking.
Bold direction change → ask the user (decision inbox). Routine refinement → auto.

## Phase 3: the quality cascade (in this order)

Run each via `deepen` with progressive deepening; fix every pass's
findings under the green ratchet before the next pass.

```
1. run 3 UX audits on the app as a whole (ux-scan), fixing what each surfaces
2. then run 5 UI polishes on the key screens (polish-pass)
3. then run 3 design-intel passes for design intelligence (styles, palettes,
   type, layout, components)
```

Default counts 3 / 5 / 3 are the calibrated starting points, tunable per project.

## Phase 4: exhaust until complete

```
keep going until the passes stop finding material improvements. if a genuinely new
premium DIRECTION shows up, re-run the wow idea-forge prompt on that (re-check the
late-stage guard first) instead of grinding the old direction. then show me the
before/after for final sign-off.
```

Stop on **saturation** (diminishing improvements per pass/token), not literal zero:
the same convergence logic as project-autopilot's CONVERGENCE.md.