Same prompt, both times

Before and after tastemaker

The prompt was build a landing page for a coffee subscription, unchanged, in both panels. Left is what a plain agent produces with no skill installed. Right is the same request with tastemaker installed: a generated Elegant/editorial palette, real fetched icons, a constructed mark, and motion, in one pass.

Before — no skillOpen full page →

After — with tastemakerOpen full page →

What actually changed, and why

Palette: indigo-to-purple gradient (the single most recognizable "AI made this" signal) vs. a palette generated fresh for this project's mood, verified against the contrast contract before it shipped.
Logo: a letter in a gradient box vs. a constructed geometric mark (a coffee bean), the same rule as any letter-in-a-box logo the checklist flags.
Icons: emoji standing in for icons vs. real fetched icons from a set matched to the mood, tinted to the accent.
Sections: three identical cards, heading plus two sentences vs. a visual step-by-step, an asymmetric lead feature with a real stat, and a stat row, show-don't-tell instead of a text wall.
Motion: a fully static page vs. a sequenced hero entrance and scroll reveals, reduced-motion-aware.
Type: default line-height on a 44px headline vs. tightened leading on the display type, a real heading/body pairing instead of one font doing both jobs.

Both pages are real, live HTML, not mockups. ← back to tastemaker