Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Turn a marketing brief or product offer into a finished short-form video ad built in HyperFrames — hook, script, shot-by-shot storyboard, scene-by-scene HTML composition, captions, and voiceover. Outputs platform cuts (vertical 9:16, square, 16:9). Use when you need a launch ad, product promo, social video, or paid-creative concept ready to render.
.claude/skills/onewave-ai-hyperframes-ad-director/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 72% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 57% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 94% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 89% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -35% | 0% |
You bring the offer; this skill directs the ad. It takes a brief and produces a real, renderable HyperFrames composition — not a vague concept — with a scroll-stopping hook, a tight script, a storyboard, and the scene HTML wired for the HyperFrames render pipeline.
This skill orchestrates the build. For the mechanics of authoring compositions, captions, transitions, and TTS, invoke the hyperframes skill (and hyperframes-cli for init/lint/preview/render, hyperframes-media for voiceover/background-removal). This skill owns the creative direction and the end-to-end assembly.
Pull or ask for:
icp-deep-scanner output if available, so the hook speaks to a real prospect.claude-design-system-architect token set exists, reuse it for visual consistency.Honor brand rules: no purple and no emoji in any visual output unless the brand kit explicitly calls for them; prefer the brand's own palette and clean iconography.
Scaffold and author the HyperFrames project (via the hyperframes / hyperframes-cli skills):
edge-tts with a neural voice if the user prefers); keep timing aligned to captions.Lint and preview (npx hyperframes lint / preview) before declaring it done. Never claim it renders without running the pipeline.
Produce the platform variants the brief needs: 9:16 vertical, 1:1 square, 16:9. Note any per-platform trims (vertical needs the hook framed for thumb-stopping; pre-roll needs the brand in the first 5s).
markdown# Ad: {Product} — {platform} - Hook options (3) + chosen - Full script - Storyboard table - Composition path + render command - Platform cuts produced - Suggested A/B (hook variant to test)
Offer to pressure-test the hook and script through prospect-panel-simulator before spending on media, and to spin a longer-form demo via hyperframes-sales-demo-builder.
The hook earns the most effort · captions always (muted autoplay) · reuse the brand kit, no purple/no emoji unless branded · lint + preview before claiming done · ground the message in a real persona.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 28,013 | 37,503 | +34% | 1 | 1 | 0% | 5,560 | 8,927 | +61% | 0 | 0 | — |
case-02 | fail→fail | 40,285 | 35,748 | -11% | 1 | 1 | 0% | 8,287 | 8,533 | +3% | 0 | 0 | — |
case-03 | fail→fail | 28,428 | 35,705 | +26% | 1 | 1 | 0% | 5,374 | 8,778 | +63% | 0 | 0 | — |
case-04 | fail→pass | 17,417 | 21,497 | +23% | 1 | 1 | 0% | 2,786 | 4,797 | +72% | 0 | 0 | — |
case-05 | pass→pass | 16,210 | 13,465 | -17% | 1 | 1 | 0% | 2,552 | 2,999 | +18% | 0 | 0 | — |
case-06 | fail→fail | 11,151 | 7,484 | -33% | 1 | 1 | 0% | 1,777 | 2,035 | +15% | 0 | 0 | — |
case-07 | fail→pass | 13,285 | 14,553 | +10% | 1 | 1 | 0% | 2,227 | 3,492 | +57% | 0 | 0 | — |
case-08 | fail→fail | 11,363 | 5,337 | -53% | 1 | 1 | 0% | 1,843 | 1,666 | -10% | 0 | 0 | — |
case-09 | fail→pass | 5,996 | 7,237 | +21% | 1 | 1 | 0% | 978 | 1,899 | +94% | 0 | 0 | — |
case-10 | pass→pass | 13,157 | 14,990 | +14% | 1 | 1 | 0% | 2,218 | 2,905 | +31% | 0 | 0 | — |
case-11 | pass→pass | 8,627 | 9,217 | +7% | 1 | 1 | 0% | 1,316 | 2,319 | +76% | 0 | 0 | — |
case-12 | fail→fail | 9,895 | 2,544 | -74% | 1 | 1 | 0% | 1,405 | 1,319 | -6% | 0 | 0 | — |
case-13 | fail→pass | 20,694 | 5,077 | -75% | 1 | 1 | 0% | 944 | 1,783 | +89% | 0 | 0 | — |
case-14 | fail→pass | 14,226 | 3,710 | -74% | 1 | 1 | 0% | 2,146 | 1,392 | -35% | 0 | 0 | — |
case-15 | fail→pass | 15,404 | 15,537 | +1% | 1 | 1 | 0% | 2,479 | 3,369 | +36% | 0 | 0 | — |
case-16 | pass→fail | 14,166 | 10,046 | -29% | 1 | 1 | 0% | 2,404 | 2,470 | +3% | 0 | 0 | — |
case-17 | pass→pass | 12,905 | 10,346 | -20% | 1 | 1 | 0% | 2,142 | 2,414 | +13% | 0 | 0 | — |
case-18 | fail→pass | 13,330 | 7,795 | -42% | 1 | 1 | 0% | 2,075 | 2,226 | +7% | 0 | 0 | — |
case-19 | pass→pass | 12,349 | 10,181 | -18% | 1 | 1 | 0% | 2,184 | 2,672 | +22% | 0 | 0 | — |
case-20 | pass→pass | 9,800 | 3,355 | -66% | 1 | 1 | 0% | 1,504 | 1,435 | -5% | 0 | 0 | — |
case-21 | fail→fail | 17,616 | 15,900 | -10% | 1 | 1 | 0% | 3,003 | 3,424 | +14% | 0 | 0 | — |
case-22 | fail→fail | 39,061 | 46,021 | +18% | 1 | 1 | 0% | 8,240 | 9,122 | +11% | 0 | 0 | — |
case-23 | pass→pass | 6,588 | 17,435 | +165% | 1 | 1 | 0% | 1,032 | 3,748 | +263% | 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. 23 cases were attempted, and 22 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +26 percentage points is the difference between those two pass rates over the 22 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
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.