agentleFS
wald21x6-opswald21x6-ops/builder-kit/skills/website/SKILL.md

website

Skill1 starsChanged 23 days agoMITView on GitHub

Playbook for websites: business sites, landing pages, portfolios, agency sites, redesigns, clones of a reference site, and client revision rounds. Adds competitor UX teardowns, one measured reference per section, a hero-first section build loop, phone and laptop designed together, real working forms, and revision rules that change only what was asked. Use it with the build skill whenever the job is a website or web page, including "make it like this site", "redesign my site", "the client sent changes", "fix the mobile version", "make this section look like X".

10 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

7 rules. 0 are common, 0 are distinctive, and 7 are very specific to this project.

  • 0

    Common

    Most of the 7 edited versions we read keep these.

  • 0

    Distinctive

    This author's own choices. Worth a look.

  • 7

    Very specific

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

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

    Use this with build. It adds website-specific steps to each stage.

    line 9
  2. Very specific:

    Do a competitor UX teardown and pick one reference per section (see research/references/sources.md → Website).

    line 17
  3. Very specific:

    Follow references/section-loop.md (capture → drive the live reference → build in your tokens → check both sizes → compare side by side → …

    line 37
  4. Very specific:

    Assets: use the user's real assets first.

    line 49
  5. Very specific:

    Run page-check.js at 390×844 and 1366×768 on the live URL, along with:

    line 54
  6. Very specific:

    Follow references/revisions.md.

    line 62
  7. Very specific:

    Use references/launch.md, and put everything in LAUNCH.md as it comes up:

    line 67

“Common” is measured: kept by at least 6 in 10 of the 7 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

Website playbook

Use this with build. It adds website-specific steps to each stage. <kit> = the Builder Kit folder named in the session-start notice (usually ~/.claude/skills/builder-kit).

Target (extra questions)

What makes a customer trust this business? Where do visitors come from (Google, Instagram, a link sent on WhatsApp)? That decides phone-first and how much SEO matters. What real material exists (logo, photos, prices, text)? Who approves the look: the user, or a client? Write the answers into BRIEF.md.

Research (extra)

Do a competitor UX teardown and pick one reference per section (see research/references/sources.md → Website). The user approves the references. If profile.md says "Design references: I choose them", don't find or fetch any yourself: ask them for references and stop until you have them. Then measure them, don't just describe them: run shot.mjs --eval on the live reference to read real fonts, sizes, spacing and timings.

Design before code

  1. Concept first, components later. From the research, propose one recommended direction, argued in a few lines, plus at most two alternatives described in a sentence each. Don't build three finished sites to choose from: two sites built from a menu ended up with the same bones under different paint.
  2. Write DESIGN.md before any page code. It holds:

    • colours and type scale as named tokens, and spacing
    • the section list, each with its reference and measured values
    • the motion plan
    • a phone plan per section (keep / re-express / drop)
    • the asset plan (real vs. placeholder)

    Show it on the Task Card.

  3. Phone and laptop are designed together, never "mobile later". Most visitors arrive on phones. In the data, mobile passes done afterwards silently dropped signature features, and the user had to ask for them back.

Build: one section at a time, hero first

Follow references/section-loop.md (capture → drive the live reference → build in your tokens → check both sizes → compare side by side → look at the screenshot). Mark anything you didn't observe on the reference as "inferred", never invented.

Defaults unless the brief says otherwise:

  • A static site (plain HTML/CSS/JS or Astro), hosted free on Vercel or Cloudflare Pages.
  • Next.js only if the site needs app features.
  • Animation through CSS or GSAP, with reduced-motion respected.

Forms must arrive somewhere real. Wire them to an inbox, a Google Sheet or a database, and submit one test for real to prove it lands. A "contact form" that sent nothing was the headline finding in one audit. A mailto: form failed on phones in another.

Assets: use the user's real assets first. Borrowed images or video from a reference are allowed only as labelled placeholders (data-placeholder), each listed in LAUNCH.md as a blocker until it's replaced. Never invent reviews, client logos or numbers. If the user supplies figures, mark them "from user" in DECISIONS.md.

Prove it (website extras)

Run page-check.js at 390×844 and 1366×768 on the live URL, along with:

  • screenshots at both sizes, with 4–5 frames through any scroll animation
  • the real form submission arriving
  • an independent critic who sees only the screenshots: "What does this company sell? What would you click?"

A hero that Claude's own checks approved was later called "very confusing", and a stranger's three-second read would have caught it.

Revision rounds (client changes)

Follow references/revisions.md. The core rule is change only what's listed. First echo every requested change back as a table (quote · where on the page · what you'll do). After a change, show one section before/after, then roll it out. Unrequested "improvements" were reversed within hours, every time, in the data.

Launch

Use references/launch.md, and put everything in LAUNCH.md as it comes up:

  • domain connected
  • placeholders replaced
  • the form tested live
  • privacy page if there's a form or login
  • basic SEO (title, description, share image)
  • analytics if wanted
  • the client's written OK

Domain and DNS steps are in backend-login/consoles/.

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.