Prompt archive
Direction, not mystique.
One original portrait was generated for MOLT with Codex built-in image generation. There were no input images and no real person was referenced. The output was curated for its cloth occlusions, anonymous silhouette and negative space, then integrated with original masks and motion.
Inspiration-sweep prompt
Read-only research task following weblify-design spoor S S1. Use Firecrawl search/scrape if available to identify 4-5 current award-level sites or studios relevant to three distinct axes: organic/biotech WebGL, radical kinetic editorial typography, tactile/material luxury object presentation. Extract only design-language insights (type scale, color logic, motion/pacing, interaction ideas), not copied assets/code. Provide URLs and distilled takeaways suitable for persisting as a sweep. Do not edit files.
Complete build and motion prompt
Continue now in C:\weblify-projects\molt. Locate the image you just generated under the built-in image tool's output path, copy the selected artifact into this project under public/media/ with a stable descriptive filename, and visually inspect it before integration. Then read this repo's AGENTS.md, INDEX.md, README.md and complete project-local weblify-design, weblify-playbook and frontend-design SKILL.md files plus blueprint/spoor-S/workflow/assets/design-review. Build MOLT · An Impossible Fashion Film as Signature stand 0. Direction: a human-centred couture organism whose generated portrait tears into cloth ribbons and reweaves through shallow 3D space; typography passes behind the subject via masks/occlusion. Do not build another linear video frame-scrub or museum catalogue. Palette bone, lacquer red, bruise lilac, oxidized green, wet black. First create 3 concepts in DESIGN.md, choose the boldest, lock exact tokens/risk contract/assets/provenance and commit before code. Implement one-page experience plus /guide, /prompts, /privacy and branded 404, original masks/SVG/CSS/canvas or WebGL motion using the generated asset, self-hosted adventurous type, CSP/reduced-motion/no-media/no-JS fallbacks, a11y and visible AI-assistance disclosure. Archive the final image prompt and provenance. Do 3 Signature iteration passes and record them. Run npm ci, lint, check, build, qa:quick; inspect 1440/390 screenshots if possible and fix Blocker/High. Commit milestones. Do not deploy. Report checks/open issues/files and exact saved asset path.
Design prompt
Build MOLT · An Impossible Fashion Film as Signature stand 0. Direction: a human-centred couture organism whose generated portrait tears into cloth ribbons and reweaves through shallow 3D space; typography passes behind the subject via masks/occlusion. Do not build another linear video frame-scrub or museum catalogue. Palette bone, lacquer red, bruise lilac, oxidized green, wet black.
Implementation prompt
Implement one-page experience plus /guide, /prompts, /privacy and branded 404, original masks/SVG/CSS/canvas or WebGL motion using the generated asset, self-hosted adventurous type, CSP/reduced-motion/no-media/no-JS fallbacks, a11y and visible AI-assistance disclosure. Archive the final image prompt and provenance. Do 3 Signature iteration passes and record them.
Final image prompt
Use case: stylized-concept Asset type: vertical website hero visual for the MOLT haute-couture showcase Primary request: Create one original impossible editorial fashion sculpture: an anonymous human silhouette almost completely enveloped by immense architectural cloth ribbons that fold, bridge, coil, and cantilever around the body. The image must feel like couture becoming inhabitable architecture, with sophisticated shallow three-dimensional depth, pronounced foreground/midground/background overlap, and strong occlusion opportunities for later web typography and scroll choreography. Scene/backdrop: abstract high-fashion studio set, seamless and minimal, with clean negative areas around parts of the silhouette; no identifiable location. Subject: a non-specific, non-recognizable human silhouette; no resemblance to any real person; face obscured or abstracted by the sculptural garment. Style/medium: museum-grade editorial fashion photography blended with hyperreal sculptural 3D, high-art magazine quality, refined material realism, uncanny but elegant rather than grotesque. Composition/framing: vertical portrait composition; full or near-full figure; asymmetrical couture mass with clear negative space zones; shallow spatial stage with dramatic layered occlusions; crop-safe edges for a responsive hero. Lighting/mood: dramatic controlled studio lighting, hard lacquer highlights against soft bruised shadows, cinematic falloff, tactile texture separation, mysterious and severe. Color palette: bone, lacquer red, bruise lilac, oxidized green, wet black; disciplined color blocking, no rainbow drift. Materials/textures: architectural woven cloth, lacquered textile edges, dense matte folds, occasional wet-black reflective surfaces; convincing tension, weight, creases, and seam-free sculptural construction. Constraints: one figure only; original design; no logos; no readable text; no watermark; no brand marks; no recognizable person; preserve clean negative areas; fashion-forward, nonsexualized. Avoid: conventional runway dress, angel wings, floral motifs, fantasy armor, sci-fi bodysuit, generic flowing-fabric stock imagery, excessive particles, busy environment, collage typography, duplicated limbs, exposed face, gore.
Provenance
- Generated on 13 July 2026.
- Tool path: Codex built-in image generation.
- External source images: none.
- Recognizable real person: none.
- No exclusivity is claimed for purely AI-generated elements.
Prompt coverage
The design brief, implementation brief and final image-generation prompt above are the complete generative prompt set used for the public experience. Component code, masks, copy and QA fixes were authored directly from those briefs; no undisclosed code-generation prompt or external visual prompt was used.
Independent review prompt
Read-only adversarial design review for C:\weblify-projects\molt. Read DESIGN.md and visually inspect references/review/final-desktop-hero.png, final-desktop-reweave.png, final-mobile-hero.png, final-mobile-reweave.png, plus the anatomy screenshots available during pass 1. Score concept-truth against DESIGN.md, not reference fidelity. Use exact rubric columns: composition, color-energy, typography, illustration, motion. Rows: opening hero desktop, reweave desktop, opening hero mobile, reweave mobile, anatomy/supporting narrative. Score each 1-5 with one-sentence evidence. Gate average >=4.0 and no cell <3. Report Blocker/High/Medium/Nitpick; be strict. Do not edit files.
Release and deployment prompt
Release/deploy MOLT autonomously in C:\weblify-projects\molt. Read global Cloudflare cloudflare + wrangler SKILL.md files before Wrangler. Work only here. Audit ITERATIONS/review logs contain 3 substantive post-build passes; audit /prompts includes every design/code/image prompt, especially exact GPT Image hero prompt/provenance; audit /guide and add concise reproducible Cloudflare Pages steps: npm ci/build, wrangler whoami, project create/production branch, pages deploy dist, live QA/update flow. Keep site-specific, no secrets. Make QA stack honor QA_PORT=4332 consistently in playwright.config.ts, qa/specs/site.spec.ts and scripts/qa-lighthouse.mjs. Commit docs/harness. Run npm ci, lint, check, build, full QA_PORT=4332 npm run qa:release without QA_BASE_URL; fix real failures. Confirm molt Pages project absent, create with production branch master, deploy dist with branch master and commit metadata. Capture production URL. Run live QA_BASE_URL=<url> npm run qa:live, verify routes/404/console, normal+reduced/no-JS/no-media and live Lighthouse 3-run median; fix/redeploy until gates green. Update INDEX/README with URL/results, commit and redeploy final changes. No user questions, no other repos, no credentials in output. Return final URL, deployment ID if available, commits, local/live results, clean worktree.
Durable source copy: this public ledger. Repository provenance is archived in references/provenance/prompt-ledger.md and the image-specific record inreferences/provenance/molt-hero.md.