awwwards-singularity-studio

An elite squad for award-tier landing pages and cinematic web experiences. An end-to-end pipeline with 8 specialised agents on Squad Protocol v6, a structure argued from every brief, a build proven from rendered pixels, and QA aligned with the real Awwwards jury criteria.

What is special

• No default skeleton. Every page starts from an experience brief (journey, feeling curve, the one peak, the sentence a visitor would tell a friend) and a page grammar chosen for that brief from eight (filmic one-shot, chaptered editorial, live surface, continuous world, typographic poster, gallery or catalog, split stage, rhythmic cutlist). Navigation, opening, sequence and close follow from the grammar, and filmic one-shot has to prove the other seven wrong.
• One signature move per page, coded for that page, tested against "could someone mistake it for a kit effect?".
• A per-project fingerprint gate. Before anything is made, the plan's six structural dimensions are checked against the project's earlier deliverables (

  <project>/.nirvana/fingerprints.jsonl

  ): one client never receives the same skeleton twice. A brand's canon keeps its look; the structure is what varies.
• Metaphor first. The strategy spine runs from the Onliness Statement to one physical visual metaphor, before code.
• The client's real assets first, generated imagery only for the gaps, one style preamble, photographic by default.
• Native scroll: no Lenis, no inertial scroll. CSS scroll-driven animation first, GSAP ScrollTrigger always scrubbed, a tested scrub playhead for video.
• Proof from pixels.

  lib/render-verify

  scrolls the served page in real Chrome at desktop, phone, compact and reduced-motion sizes and fails the build on unreadable contrast over moving media, frozen clips, dead scroll, obscured focus or overflow. Pass flags are derived from its reports, never declared.
• An honest jury score: Design 40%, Usability 30%, Creativity 20%, Content 10%, a house bar of 7.0 that reopens the build instead of inflating the score, and a cold feel check before the reviewer reads the brief.
• Single-file HTML production-ready by default; the current Next.js App Router when the brief asks for a multi-route app.

Pipeline (8 phases)

1. Strategist     → Experience brief, Onliness to metaphor, grammar, signature move, fingerprint check, stack
2. Researcher     → Award-winning references (the mechanism, never the structure), trends of the last 18 months
3. Copywriter     → Real copy shaped by the grammar and each beat (zero lorem ipsum, zero invented figures)
4. Designer       → OKLCH tokens, named vibe, anti-slop, taste floor, section score, layer contract
5. Illustrator    → Client assets first, generated gaps in one visual language (Nano Banana Pro → Flux → image2-virtuoso)
6. Developer      → Build: devices on native scroll, WebGL (Three.js current stable), GSAP + easeReverse (floor 3.15), CSS scroll-driven, scrub playhead, verification hooks
7. Optimizer      → render-verify evidence, vitals budget, WCAG 2.2 AA, extractability
8. Reviewer       → Cold feel check, Awwwards score (40/30/20/10) with a house bar of 7.0, final report, registry append

Quick start

# Full pipeline: from the brief to the Awwwards score
nrv run --squad awwwards-singularity-studio:design.full_experience_pipeline.execute

# Discovery only: brief, strategy, metaphor, structure, stack, references
nrv run --squad awwwards-singularity-studio:design.discovery_flow.execute

# Copy and design only, in parallel, from an approved brief
nrv run --squad awwwards-singularity-studio:copy.design_copy_parallel.execute

# Build and optimization only, from approved copy and design
nrv run --squad awwwards-singularity-studio:design.build_pipeline.execute

# QA scoring only, on a build that already exists
nrv run --squad awwwards-singularity-studio:design.qa_gate.execute

Without :capability the router picks among the eight by the brief. Name one when you already know.

Capabilities

Every capability declares acceptance[]: what the judge checks in the run, not what the author hoped for.

Libraries

The method lives in shared libraries copied into

lib/

lib/experience-craft/

scripts/fingerprints.ts

lib/render-verify/

lib/scroll-runtime/

lib/design-intelligence/

lib/seo-geo-aeo/

scripts/audit-search-visibility.ts

lib/experience-craft/

lib/render-verify/

lib/scroll-runtime/

Experience modes

The mode is the stack; the grammar is the structure. Any mode can carry any grammar the brief earns.

Scoring thresholds

7.0 is this studio's bar, not Awwwards'. The official Honorable Mention cut is read on the Awwwards evaluation page during each run.

Requirements

Bun. The installed Google Chrome (bundled Chromium cannot decode H.264), a full ffmpeg build, and playwright-core and sharp in the shared store (nrv deps install playwright-core sharp);

bun --no-install lib/render-verify/scripts/doctor.ts

lib/scroll-runtime/device-diag.html

Author

gutomec (@gutomec)

License

SUL-1.0 (see license in squad.yaml). The third-party code in

lib/