Ship the browsable mosaic document as an isolated HTML file and document the build intent in WHAT.md so the categorization work has a concrete UI target.
Co-authored-by: Cursor <cursoragent@cursor.com>
Restrained mode now reads the palette like screen 02: primary-tinted
headline, tertiary eyebrow, secondary gradient hero, outlined ghost
CTA, and four-role gallery accents. Artboards lift with a soft shadow;
the phone gains surface bars and a second copy line. Each strategy
block sets matching image-gradient stops so committed, full palette,
and drenched transitions stay depthful instead of flat fills.
Co-authored-by: Cursor <cursoragent@cursor.com>
The heading family name sits big in its own face; one sentence in the
body face names its partner in bold and carries the why line. The
sentence demonstrates the body font, so the Heading/Body labels go.
Co-authored-by: Cursor <cursoragent@cursor.com>
Screen 04 drops the kicker and the text eyebrow, renders six large
scrollable specimen cards (heading and body each labeled and set in
their own face), and reuses screen 03's dual desktop-plus-phone
artboards with both hero copy blocks typeset live. The type hero
widens its copy column so wide display faces clear the image block.
Co-authored-by: Cursor <cursoragent@cursor.com>
Screen 03 lets operators compare restrained through drenched strategies on the dual-artboard preview; visual-cues now composes fonts.json before launch, and the picker server exposes it.
Co-authored-by: Cursor <cursoragent@cursor.com>
Show landing and app wireframes side by side with palette-driven accents instead of PRODUCT.md register and /context.json, matching v4's per-surface modes. Assisted-by: Cursor AI agent.
Co-authored-by: Cursor <cursoragent@cursor.com>
Post-rebase repairs after main moved 226 commits (site/ split to the
private impeccable-site repo, DESIGN.md spec now eight sections, register
extraction replaced by extractSectionValue, seed mode rewritten):
- document.md: transplant this branch's seed pipeline (assets, interview,
visual cues, picker) onto main's rewritten file; adopt the canonical
section vocabulary. The interplay with new-work's workshop is left for
PR review.
- picker: vendor kinpaku tokens/kit CSS, hero art, and favicon under
picker/ (site/ no longer exists here); drop the sharp resize step and
ship the compressed JPEG directly; declare astro/sharp-free deps.
- picker-server: read the register via extractSectionValue('Register')
since extractRegister is gone from context.mjs.
- palette.mjs / utils.js / test-suites.mjs / package.json / release.mjs:
rebuild as main's version plus this branch's additions, undoing hunks
the -X theirs replay had clobbered.
- craft.md / shape.md: reset to main (this branch's net change was zero).
- progress bar: animate transform instead of width (design hook finding).
Validation: bun run build green (prose validators clean), bun run test
198/198, picker-server suite 7/7.
AI-assisted (agent-implemented, maintainer-directed).
Co-authored-by: Cursor <cursoragent@cursor.com>
gemini-3-pro-image is Google's strongest image model; IMAGE_GEN_MODEL still overrides for the cheaper flash tier.
AI-assisted commit.
Co-authored-by: Cursor <cursoragent@cursor.com>
Ban hex codes from image prompts to stop Nano Banana swatch strips, require crisp deep focus and set-dressed surfaces inside close product-world scenes, and add wordless and real-estate inspection retries. The Gemini wrapper now converts JPEG bytes to PNG and infers AQ.-prefixed keys.
AI-assisted commit.
Co-authored-by: Cursor <cursoragent@cursor.com>
Ship image-gen.mjs with bfl and gemini backends plus custom delegation, document the keyless flow in image-api.md, and tighten seed Step 4 so missing keys ask once and opt-out falls back to text-only seeding. Compile now removes in-folder hero intermediates so each cue keeps one PNG.
AI-assisted commit.
Co-authored-by: Cursor <cursoragent@cursor.com>
Each specialist now generates one hero image instead of a second chroma-key sheet, and the compile step replaces crop with a simpler cues.json shape.
AI-assisted commit.
Co-authored-by: Cursor <cursoragent@cursor.com>
Drop the two-pass matte pipeline in favor of one chroma-key sheet per
cue: each specialist picks the farthest-hue key, stages larger artifacts
with shadowless crisp edges, reports CHROMA in its reply, and crop records
the hex in cues.json for browser keying downstream.
Co-authored-by: Cursor <cursoragent@cursor.com>