Agent-Skills.md

Agent Skills: Video → Scroll-Driven Landing Page

Turn any video into a cinematic scroll-driven landing page — Apple-style hero where scrolling progresses the visible frame through the video. Use when the user provides a video file and asks for "a landing page from this video", "scroll-frame website", "Apple-style scroll site", "hero that scrubs the video", "like the GitHub Copilot landing", or any equivalent. Extracts N evenly-spaced frames via ffmpeg, builds a self-contained HTML page with a sticky hero + JS scroll listener that swaps the visible frame as you scroll, plus headline, sections and CTA below. For YUV.AI projects, applies the yuv-design-system skill in Neon mode (pink/cyan/white, default for YUV.AI web) — Decks (purple/yellow) is reserved for slides only. For generic / non-YUV.AI projects, picks an appropriate palette per the source video. Output is one folder with `index.html` and a `frames/` directory — drop on any static host.

UncategorizedID: hoodini/ai-agents-skills/video-to-landing-page

Repository

hoodini/ai-agents-skills
hoodini
27963

Install this agent skill to your local

pnpm dlx add-skill https://github.com/hoodini/ai-agents-skills/tree/HEAD/skills/video-to-landing-page

Skill Files

Browse the full folder contents for video-to-landing-page.

Download Skill

Loading file tree…

skills/video-to-landing-page/SKILL.md

Skill Metadata

Name
video-to-landing-page
Description
Turn any video into a cinematic scroll-driven landing page — Apple-style hero where scrolling progresses the visible frame through the video. Use when the user provides a video file and asks for "a landing page from this video", "scroll-frame website", "Apple-style scroll site", "hero that scrubs the video", "like the GitHub Copilot landing", or any equivalent. Extracts N evenly-spaced frames via ffmpeg, builds a self-contained HTML page with a sticky hero + JS scroll listener that swaps the visible frame as you scroll, plus headline, sections and CTA below. For YUV.AI projects, applies the yuv-design-system skill in Neon mode (pink/cyan/white, default for YUV.AI web) — Decks (purple/yellow) is reserved for slides only. For generic / non-YUV.AI projects, picks an appropriate palette per the source video. Output is one folder with `index.html` and a `frames/` directory — drop on any static host.

Video → Scroll-Driven Landing Page

Build a cinematic landing page where the hero is a video playing on scroll instead of time — like Apple product pages or the GitHub Copilot landing. Frames are extracted from the source video and swapped as the user scrolls; below the sticky hero, normal-scroll sections carry copy and CTA.

When to invoke

  • User drops a video and asks for a landing page, scrolling site, scroll-frame hero, or website from a video
  • Mentions of "Apple-style scroll", "GitHub Copilot landing", "scrub-on-scroll", "frame on scroll"
  • Hebrew equivalents: "דף נחיתה מסרטון", "אתר עם גלילה", "סרטון בגלילה"

Save location

Default: ~/Documents/yuv-projects/landings/<slug>/ — always save landing pages here so you can find them again. Create the parent directory if missing.

mkdir -p ~/Documents/yuv-projects/landings
cd ~/Documents/yuv-projects/landings
# Build <slug>/ here.

Final path: ~/Documents/yuv-projects/landings/<slug>/. Tell the user where the landing lives at the end of every build.

Workflow (8 steps)

  1. Probe the source — ffprobe for duration, fps, dimensions, audio.
  2. Pick frame count — default 80. Floor 40 (choppy if lower), ceiling 160 (heavy page). Choose by duration: ≤10s → 50, 10–30s → 80, 30–60s → 120, >60s → 160.
  3. Decide output directory — default ~/Documents/yuv-projects/landings/<slug>/. The <slug> is short, derived from the video name or the topic. Override only on explicit request.
  4. Extract frames — run references/extract-frames.py <source-video> <output-dir> [count]. Outputs output-dir/frames/f_0001.jpg… with consistent JPG quality 82 and a max-width of 1920 (downscaled if source is larger).
  5. Copy and customise references/landing-template.html into output-dir/index.html:
    • Replace the __FRAMES__ placeholder with <img> tags for each frame (the build script does this for you when you pass --build-html).
    • Fill in __HEADLINE__, __TAGLINE__, __CTA_TEXT__, __CTA_HREF__, sections.
    • For Hebrew content: set <html lang="he" dir="rtl"> and swap fonts to Rubik/Assistant (see references/design-notes.md).
  6. Optimise frames — the build script already scales to 1920px max-width. If the page total weight is > 12 MB, drop quality to 75 or reduce frame count.
  7. Preview — python -m http.server inside the output directory and open http://localhost:8000. Scroll feels buttery on a decent machine.
  8. Deploy — drop the folder on Vercel/Netlify/Cloudflare Pages. No build step needed.

The scroll mechanic

  • Outer scroller .scroll-stage height set in viewport units (default 600vh = scroll 5 screens to traverse all frames).
  • Sticky hero position: sticky; top: 0; height: 100vh containing all frames absolutely positioned and hidden except the active one.
  • Scroll listener computes progress = scrolledIntoStage / scrollableHeight clamped 0–1.
  • idx = floor(progress * (N - 1)) — swap the .active class to that frame.
  • Wrap the listener in requestAnimationFrame to keep 60fps.

Design rules

  • If the project is YUV.AI-branded (Yuval personally, his portfolio, his brand): call yuv-design-system and lock it to a mode by medium —
    • Neon mode (DEFAULT for YUV.AI web): hot pink (#FF1464) + neon cyan (#00E5FF) + white or rich black canvas. Anton + Inter (English) / Rubik + Assistant (Hebrew). 0px or 999px radii.
    • Warm Editorial mode: only for Hope, Marcus, bigcats.ai, practical.yuv.ai — pink + yellow + bone, off-white never pure #fff.
    • Never use Decks mode (Fly High purple/yellow) for a landing page — that palette is reserved for slide decks only.
  • If the project is not YUV.AI-branded: do not apply yuv-design-system. Pick a palette that fits the source video's mood. Anton + Inter remain reasonable defaults but aren't mandatory.
  • Hero overlays should be subtle — a vignette and a single big headline; don't drown the frame.
  • Below-hero sections: light canvas (white in Neon mode, off-white in Warm Editorial, bone if going generic-warm), large Anton heading, body copy in Inter at 18–22px, CTA as a sharp pill (border-radius: 999px) — pink fill in Neon mode, brand-appropriate color otherwise.
  • For Hebrew: respect RTL, use Rubik 900 for headlines, Assistant 400/500 for body.

File references

| File | Purpose | | --- | --- | | references/extract-frames.py | ffmpeg-based frame extractor; optionally builds the full HTML in one go | | references/landing-template.html | Self-contained landing-page skeleton with the scroll mechanic | | references/design-notes.md | Typography, palette, and motion defaults |