agentleFS
duncan-buildroomduncan-buildroom/profit-room-skills/skills/build-site/SKILL.md

build-site

Skill18 starsChanged 4 months agoMITView on GitHub

Build a $10k-class animated website for the user's brand. Interviews them, picks a scroll style, designs to a strict set of design laws, generates or sources bespoke hero media, and ships a single premium animated landing page, product site, or portfolio. Use when the user wants a high-end animated site, a scroll-scrub product page, or a portfolio that looks expensive.

2 other projects use a version of this.

Adopted by

Most recently taken up by

The file

Show its rules and text

We split this skill into the rules it gives an agent

13 rules. About 0 are common, 11 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈11

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If it doesn't exist, run the personalize skill first (or just tell me your name, niche, offer, and audience — I'll create it).

    line 13
  2. Distinctive:

    If a field this skill needs is blank, I'll ask one or two quick questions and save the answers back to the profile so you only answer once.

    line 14
  3. Distinctive:

    Never invent facts or results about the user — use only what's in the profile, or ask.

    line 15
  4. Distinctive:

    Every great site here = the design laws (the quality floor) + a unique config + unique assets (the customization).

    line 19
  5. Distinctive:

    Do you want bespoke AI-generated media, or should we use stock/placeholder media for now?

    line 29
  6. Distinctive:

    Do not over-ask. Pull what you can from the brand profile, infer the rest, and propose.

    line 31
  7. Very specific:

    Read references/scroll-catalog.md.

    line 34
  8. Distinctive:

    Keep the asset budget: the spine uses 2 videos + 3 images, each exactly once.

    line 51
  9. Distinctive:

    With generated media: use your available image/video generation (see Requirements) to render the 2 videos + 3 images from the prompts, chop …

    line 68
  10. Distinctive:

    Only re-render the asset that changed.

    line 79
  11. Very specific:

    No em dashes in copy. No #000/#fff.

    line 83
  12. Distinctive:

    If a render's START and END drift into a different object (exploded/transform styles), fall back to a loop — consistency over spectacle.

    line 86
  13. Distinctive:

    No API keys are embedded in this skill.

    line 92

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “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

Build Site

You are building the user a premium animated website that looks like it cost $10,000. The quality comes from two things working together: a strict set of design laws (the quality floor) wrapped around a unique config and unique assets (what makes it theirs). You never lower the floor; you make the site specific.

Personalization

This skill builds a site in your brand voice and look. Before running, load the brand profile at ~/.claude/brand-profile.md.

  • If it doesn't exist, run the personalize skill first (or just tell me your name, niche, offer, and audience — I'll create it).
  • If a field this skill needs is blank, I'll ask one or two quick questions and save the answers back to the profile so you only answer once.
  • Never invent facts or results about the user — use only what's in the profile, or ask.

The one idea

Every great site here = the design laws (the quality floor) + a unique config + unique assets (the customization). The look is guaranteed by obeying the laws; you make it specific with the brand's palette, copy, and bespoke renders. Read both references before you start:

  • references/design-laws.md — palette, type, copy, and layout laws. Obey them.
  • references/scroll-catalog.md — the seven scroll styles (A–G), the render laws, and the frame-sequence technique that makes scrubbing feel premium.

Flow

1. Interview (keep it to 4 questions)

  • What is the product or brand, in one line?
  • Who is it for, and what should they feel?
  • What is the one thing they should remember?
  • Do you want bespoke AI-generated media, or should we use stock/placeholder media for now?

Do not over-ask. Pull what you can from the brand profile, infer the rest, and propose.

2. Pick the scroll style

Read references/scroll-catalog.md. Match the product to a style — do not default to one:

  • A Ambient loop — a single hero object that just looks beautiful (spirits, fragrance, a watch). No scroll effect.
  • B Scroll-scrub — a material or product that transforms/rotates as you scroll (most products, textures).
  • C Cursor-spotlight — something that hides in the dark and is revealed (lighting, a car, anything dramatic).
  • D Horizontal pan-lock — a place or a lineup you move sideways through (a workshop, a coast, a collection).
  • E Exploded — a built object worth taking apart (a watch, earbuds, a camera, anything mechanical).
  • F Push-through — a space you fly into (architecture, an experience, an interface).
  • G Scrollytelling — a process worth watching unfold (coffee, forging, making anything).

State your pick and why in one line.

3. Design the config (this IS the design work)

Decide every design choice up front, obeying references/design-laws.md:

  • Palette: OKLCH only, never #000/#fff, one committed accent. Choose a theme from a one-sentence scene, not from the category.
  • Type: a distinctive display face + a clean grotesk body. Real hierarchy (≥1.25 scale ratio).
  • Copy: tight, specific, no filler. No em dashes. Headlines short.
  • Captions (3 short beats): 3–5 words each, one accent word each.
  • Keep the asset budget: the spine uses 2 videos + 3 images, each exactly once. Never add more slots; never reuse an asset. That restraint is what keeps it clean.

4. Write the render prompts (or plan placeholders)

Following the render laws in references/scroll-catalog.md:

  • 2 videos: hero (the header move) + middle (a different second video that matches the hero's light and palette — a sibling prompt, not the same shot).
  • 3 images: a product/context shot, a macro, and a hero still.
  • Lock the camera, constant velocity, motion blur off. For any scroll-scrubbed style, render long and export a WebP/PNG frame sequence — that is what makes scrubbing feel premium.

5. Build the page

Build a single self-contained site using free, open-source libraries (pull from a CDN or npm):

  • GSAP + ScrollTrigger for pin + scrub.
  • Lenis for smooth inertia scrolling.
  • A canvas flipbook driven by scroll progress for any frame-sequence hero (Style B–G).
  • SplitType for caption choreography.

Follow the canonical section spine in references/scroll-catalog.md (hero → thesis → feature pairs → gallery band → second moment → editorial rows → CTA → footer). Enforce the universal legibility rules (dark halo behind hero text, translucent content sections over a slow mesh-gradient background, a seam gradient so hero footage fades into content).

  • With generated media: use your available image/video generation (see Requirements) to render the 2 videos + 3 images from the prompts, chop scroll videos into a frame sequence with ffmpeg, and wire them in.
  • Without: drop in tasteful placeholders (a solid brand-tinted plate or a simple CSS loop) into the exact same slots so the structure and motion are correct, and mark each slot so the user can swap real media in later.

Serve the folder locally and open the page. Show the user.

6. Iterate by conversation

This is the whole point. The user says it, you change the design:

  • "Make it darker / warmer" → edit the palette (no re-render).
  • "Shorter headline / different copy" → edit copy (no re-render).
  • "Use the exploded style" → change the style + rewrite the hero render prompt, re-render the hero only.
  • "Swap the product shot" → rewrite one image prompt, re-render that one image. Only re-render the asset that changed. Copy and palette edits never need a render.

Hard rules (do not break)

  • 2 videos + 3 images, each used once. No gallery grids, no reused frames as filler.
  • No em dashes in copy. No #000/#fff. One accent.
  • Scroll-scrubbed heroes ship as a frame sequence, never a raw scrubbed <mp4> (raw video stutters on backward scrub).
  • The footer credit line is the brand's own (from the brand profile), not anyone else's.
  • If a render's START and END drift into a different object (exploded/transform styles), fall back to a loop — consistency over spectacle.

Requirements

  • A modern browser + a local static server (e.g. npx serve) to preview. Always available.
  • ffmpeg — to chop rendered videos into scroll frame sequences. If missing, tell the user to install it, or fall back to ambient-loop (Style A) which needs no frames.
  • Image/video generation (optional) — for bespoke hero media. Use whatever generator you have access to (e.g. an MCP image/video tool or a CLI). If none is available, build the full structure with placeholders and tell the user exactly which slots to fill.
  • No API keys are embedded in this skill. If a generator needs a key, source it yourself before running.

Result

A site indistinguishable in quality from a top awwwards build, but unmistakably the user's: their brand, their product, their words, their bespoke renders.

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.