Prompt Lab
The exhibition catalogue
Give an AI builder this brief and it returns today's system, sized for a real exhibition client.
Build a one-page museum exhibition site for a show called [EXHIBITION TITLE], running [DATES].
Art direction: a dim gallery after hours. Matte charcoal-plum ground (#18141F) with slightly lighter panels (#221C2B), alabaster text (#F1EAE0), warm stone grey for captions (#A79E92), and one flat cobalt accent (#4A5CD8) for the banner spine, rules, and annotations. Matte surfaces only: no gradients, no glow, no drop shadows. Corners square, borders 1px.
Type: Instrument Serif (regular and italic, single weight) for the poster title and all headings at generous sizes; Albert Sans for body at 19px with 1.65 line height; Courier Prime for accession numbers, wall-label captions, and metadata. Headings are never bold: scale and the serif do the work.
Layout, top to bottom:
1. Thin top bar: wordmark and two links, 1px rule below.
2. Poster hero: a solid cobalt vertical banner spine down the left edge carrying the issue or event metadata in vertical mono type; beside it, a mono kicker, the exhibition title in large serif, a two-line deck, and a dates row with a boxed "view the exhibition" button.
3. Full-width photographic installation plate in a thin border with padded matte, mono caption row beneath.
4. Didactic panel: one large annotated SVG or image chart in a bordered panel, key data points labeled in mono directly on the chart, followed by a wall-label caption card (figure number, source, one sentence of interpretation) with a 3px cobalt top rule.
5. The collection: works grouped under room headings (mono small caps), each an accession card with a mono accession number, a medium line, a serif title, two sentences, and a provenance link.
6. Plan-your-visit band: three cells (hours, admission, getting here) with cobalt top rules on a darker ground.
Readability guardrails: body no smaller than 18px, lines under 70 characters, WCAG AA contrast for every caption measured against its actual surface, descriptive link text, hover and focus states with 150ms easing, and a reduced-motion media query. Never set readable-looking text inside images; no emoji icons, no neon, no purple-blue gradients.
Runs in v0, Bolt, Lovable, Figma Make, or Beaver Builder AI; any code agent can follow it.