agentleFS
kuker24kuker24/OpenCodeBestFriend/skills/scroll-craft/SKILL.md

scroll-craft

Skill0 starsChanged 2 months agoMITView on GitHub

Build scroll-led storytelling websites where scroll is the timeline: scrollytelling, signature interaction, pin/pan/reveal, layered hero, cinematic scroll without camera flight. Use when the user runs /scroll-craft, or asks for a site where scrolling tells the product story. Not for ordinary dashboards or landings (impeccable), continuous camera fly-through / diorama / 3D world (scroll-world), chrome motion (emil-design-eng), or photoreal ads (visual-studio).

6 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

9 rules. 0 are common, 6 are distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 5 edited versions we read keep these.

  • 6

    Distinctive

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Do not read every reference on every request.

    line 27
  2. Distinctive:

    Follow the project stack. Do not force a standalone HTML file onto React/Vue.

    line 42
  3. Very specific:

    No scroll hijacking. No wheel/touch/keyboard traps.

    line 45
  4. Very specific:

    No KIE, Higgsfield, ffmpeg, Playwright, font CDN, or remote assets on the core path.

    line 46
  5. Distinctive:

    Do not invent brand facts, stats, testimonials, or user quotes.

    line 48
  6. Distinctive:

    Read PRODUCT.md, DESIGN.md, tokens, and existing assets.

    line 71
  7. Distinctive:

    Write the visitor journey and feeling curve.

    line 73
  8. Distinctive:

    Verify along the scroll: desktop, mobile portrait, reduced motion, reverse, anchor jump, keyboard.

    line 80
  9. Distinctive:

    Tier 1 (default): no generation, no video. Complete experience offline.

    line 89

“Common” is measured: kept by at least 6 in 10 of the 5 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Scroll Craft

Scroll is the timeline. The page is a story the visitor drives with their hand. One grammar, one peak, one signature move. Premium means clear composition and a meaningful interaction, not video, pinning, or effects.

Load when

NeedFile
Journey, feeling curve, peakjourney.md
Eight grammarsgrammars.md
Devices and --sc-pdevices.md
Signature + fingerprintsignature-fingerprint.md
Optional hero depthhero-depth.md
Mobile, keyboard, reduced motionmobile-accessibility.md
Scroll verificationverification.md
Continuous World / media / QAhandoff.md

Do not read every reference on every request. Engine: scrollcraft.js, scrollcraft.css. The mechanism CSS is safe for an existing design system; the scoped starter theme is optional. Copy what the project needs; do not edit the mechanism per page. Drive bespoke behaviour from --sc-p. Attribution: NOTICE.md.

Hard rules

  • One primary specialist. This skill owns the scroll story. Do not auto-load Impeccable for every build step.
  • Ordinary scrollable UI stays Impeccable. Words like premium, cinematic, interactive, or layered alone are not enough.
  • Continuous camera / fly-through / diorama / 3D world → hand off to scroll-world after a minimum brief. Do not implement worldflight here.
  • Follow the project stack. Do not force a standalone HTML file onto React/Vue.
  • Semantic HTML for headings, CTA, nav, and forms. Core content readable with JS off and with reduced motion.
  • No scroll hijacking. No wheel/touch/keyboard traps.
  • No KIE, Higgsfield, ffmpeg, Playwright, font CDN, or remote assets on the core path. Tier 1 is HTML/CSS/transforms/sticky/local assets/light JS.
  • Do not invent brand facts, stats, testimonials, or user quotes.
  • Live Surface controls must work. Label synthetic demos. Never fake a successful transaction.
  • Design V2 is optional, offline, read-only. Empty bank does not block build.
  • Store brief/fingerprint/evidence under .scratch/scroll-craft/<slug>/ in the user project. Not in this skill folder. No new global home.

Handoff

RequestOwner
Scroll-led storytelling / scrollytellingthis skill
/scroll-craft plus Continuous Worldthis skill enters, then scroll-world
Continuous camera, diorama, 3D world (no slash)scroll-world
Dashboard, ordinary landing, design systemimpeccable
Hover/press/easing after the surface existsemil-design-eng
Standalone photoreal / ads / identityvisual-studio
Deterministic HTML composition rendered to videohyperframes
Exploratory QAbrowser-act
Observed browser causeopencode-chromium-cdp then chrome-devtools-axi

Run

  1. Read PRODUCT.md, DESIGN.md, tokens, and existing assets. Reuse answers.
  2. Lock the brief. Interview only missing decisions. Mark assumptions.
  3. Write the visitor journey and feeling curve. One engineered peak.
  4. Pick one page grammar. grammars.md.
  5. Invent one domain-relevant signature interaction with a static fallback.
  6. Compare fingerprint if history exists; else record NO_HISTORY. Design V2 retrieval is optional.
  7. Assign a device per beat, with fallback. No quota of device families.
  8. Build in the project stack. Copy the engine if the page uses data-sc-*.
  9. Verify along the scroll: desktop, mobile portrait, reduced motion, reverse, anchor jump, keyboard. verification.md.
  10. Fix, save evidence under .scratch/scroll-craft/<slug>/, record fingerprint.

/scroll-craft with an explicit Continuous World brief: write the story, waypoints, and feeling curve, then hand off. Do not take the renderer.

Media

  • Tier 1 (default): no generation, no video. Complete experience offline.
  • Tier 2: existing local images with clear license. No ffmpeg.
  • Tier 3: video only if the brief needs it and native tooling exists.

A finished non-video path is normal, not DEGRADED. Use DEGRADED only when a requested capability cannot be met. Do not invent tools.

Output

The page (or mount into the existing site), the journey table, grammar and why the others lost, signature move, fingerprint (NO_HISTORY or diffs), peak, verification actually performed, and limits (emulation vs device). Do not dump prompts or claim uniqueness without evidence.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.