Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Non-animation creative direction for HyperFrames videos. Use for design spec (frame.md / design.md) handling, palettes, typography, narration, beat planning, audio-reactive visuals, composition patterns, and brand / style decisions. For atomic motion patterns and scene blueprints, use `hyperframes-animation`.
.claude/skills/coco-research-hyperframes-creative/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 15% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 17% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -21% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 49% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -16% | 0% |
Brand, pacing, style, narration, and composition direction. Use after the technical contract from hyperframes-core is in place.
For motion patterns, scene blueprints, transitions, and CSS marker effects, use hyperframes-animation — this skill is intentionally non-animation.
> Read these two FIRST for any non-trivial composition — they override web instincts: > > - references/house-style.md — "interpret the prompt, generate real content," the lazy-default list, and the background/foreground layer recipe. This is what turns a literal restyle into a _concept_. > - references/video-composition.md — video-medium scale, depth, and foreground detail. It explains how to avoid empty web-page layouts without imposing a universal element count. > > Skipping these is the single biggest cause of generic, web-page-looking output. They are not optional rows in the routing table below — for anything beyond a one-line edit, open both before you choose colors or write HTML.
frame.md → design.md → DESIGN.md) and how to parse it (frontmatter = normative, prose = context) are defined once in references/design-spec.md — resolve and load per that doc.frame-presets/ (adopt a FRAME.md as frame.md; see references/design-spec.md)references/visual-styles.mdreferences/house-style.mdreferences/design-picker.mdreferences/beat-direction.md. For scene transitions, jump to hyperframes-animation/transitions/.references/motion-principles.md (high-level guardrails), then go to hyperframes-animation for atomic rules.| Topic | Read | | ----------------------------------------------------------------------------- | ---------------------------------------------- | | Adopt a ready-made frame-preset as frame.md (optional) | frame-presets/ · references/design-spec.md | | Default palettes, motion, typography, lazy defaults to question | references/house-style.md | | Named style presets, mood-to-style routing | references/visual-styles.md | | Palette-specific color tokens | palettes/*.md | | Composition patterns — PiP, text-behind-subject, title card, slide show | references/composition-patterns.md | | Stats / infographic presentation | references/data-in-motion.md | | Structured expansion for open-ended prompts | references/prompt-expansion.md | | Video-medium density, scale, color, frame composition | references/video-composition.md | | Per-beat direction, rhythm planning, transition timing | references/beat-direction.md | | Post-authoring spec verification (colors, type, corners, spacing, depth) | references/design-adherence.md | | High-level motion guardrails and GSAP-quality rules | references/motion-principles.md | | Font selection, pairings, rendered-video type guardrails | references/typography.md | | Story doctrine — hook language, value-before-evidence, storyboard-as-proposal | references/story-spine.md | | Script pacing, tone, openings, number pronunciation | references/narration.md | | Precomputed audio bands mapped to motion | references/audio-reactive.md |
scripts/contrast-report.mjs — inspect contrast warnings from rendered frames.scripts/extract-audio-data.py — pre-extract audio bands for audio-reactive compositions.scripts/package-loader.mjs — support script for bundled creative tooling.contrast-report.mjs resolves helper packages from the current project first, then can bootstrap the bundled HyperFrames package version. Set HYPERFRAMES_SKILL_PKG_VERSION=<version> only when running the skill outside the bundled CLI/skill install and you need to pin that bootstrap version explicitly.
Run from the repo root with explicit paths, for example:
bashpython skills/hyperframes-creative/scripts/extract-audio-data.py <audio-file>
Animation analysis (animation-map.mjs) lives in hyperframes-animation/scripts/.
hyperframes-core technical rules.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | pass→pass | 11,398 | 7,819 | -31% | 1 | 1 | 0% | 1,158 | 1,737 | +50% | 0 | 0 | — |
case-02 | pass→pass | 7,171 | 3,233 | -55% | 1 | 1 | 0% | 1,159 | 1,800 | +55% | 0 | 0 | — |
case-03 | pass→pass | 18,756 | 9,152 | -51% | 1 | 1 | 0% | 2,638 | 1,986 | -25% | 0 | 0 | — |
case-04 | fail→pass | 11,689 | 4,367 | -63% | 1 | 1 | 0% | 1,503 | 1,731 | +15% | 0 | 0 | — |
case-09 | pass→pass | 8,537 | 4,090 | -52% | 1 | 1 | 0% | 1,399 | 1,773 | +27% | 0 | 0 | — |
case-05 | fail→pass | 10,205 | 4,721 | -54% | 1 | 1 | 0% | 1,646 | 1,928 | +17% | 0 | 0 | — |
case-06 | fail→pass | 17,213 | 7,629 | -56% | 1 | 1 | 0% | 1,961 | 1,549 | -21% | 0 | 0 | — |
case-07 | fail→pass | 6,020 | 7,961 | +32% | 1 | 1 | 0% | 1,077 | 1,604 | +49% | 0 | 0 | — |
case-08 | fail→pass | 17,083 | 1,994 | -88% | 1 | 1 | 0% | 1,823 | 1,537 | -16% | 0 | 0 | — |
case-10 | fail→pass | 14,525 | 7,556 | -48% | 1 | 1 | 0% | 1,976 | 1,465 | -26% | 0 | 0 | — |
case-11 | fail→pass | 11,117 | 6,734 | -39% | 1 | 1 | 0% | 906 | 1,520 | +68% | 0 | 0 | — |
case-12 | fail→pass | 15,930 | 1,952 | -88% | 1 | 1 | 0% | 1,720 | 1,457 | -15% | 0 | 0 | — |
case-13 | fail→pass | 12,707 | 7,086 | -44% | 1 | 1 | 0% | 1,140 | 1,498 | +31% | 0 | 0 | — |
case-14 | fail→pass | 12,281 | 7,527 | -39% | 1 | 1 | 0% | 1,127 | 1,512 | +34% | 0 | 0 | — |
case-15 | fail→pass | 8,853 | 7,399 | -16% | 1 | 1 | 0% | 1,303 | 1,739 | +33% | 0 | 0 | — |
case-16 | fail→pass | 15,451 | 6,955 | -55% | 1 | 1 | 0% | 1,837 | 1,409 | -23% | 0 | 0 | — |
case-17 | pass→pass | 12,557 | 8,291 | -34% | 1 | 1 | 0% | 1,334 | 1,695 | +27% | 0 | 0 | — |
case-18 | fail→pass | 10,096 | 1,950 | -81% | 1 | 1 | 0% | 1,607 | 1,524 | -5% | 0 | 0 | — |
case-19 | fail→pass | 12,141 | 2,762 | -77% | 1 | 1 | 0% | 1,985 | 1,636 | -18% | 0 | 0 | — |
case-20 | fail→fail | 33,457 | 18,081 | -46% | 1 | 1 | 0% | 3,156 | 3,902 | +24% | 0 | 0 | — |
case-21 | pass→pass | 12,212 | 7,750 | -37% | 1 | 1 | 0% | 1,941 | 1,611 | -17% | 0 | 0 | — |
case-22 | fail→pass | 10,194 | 2,453 | -76% | 1 | 1 | 0% | 845 | 1,474 | +74% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +68 percentage points is the difference between those two pass rates over the 22 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.