agentleFS
ldaumeldaume/agentic-engineering-harness/skills/design/design-web/SKILL.md

design-web

Skill3 starsChanged 3 months agoMITView on GitHub

Use whenever designing, structuring, or reviewing a website rather than an app - marketing site, landing page, homepage, editorial or magazine site, e-commerce storefront, docs site - for information architecture, layout, typography, visual direction, motion, award-style inspiration, or Core Web Vitals (Website gestalten, Landingpage-Layout, Webdesign, Seitenstruktur, Ladezeit). Not for the copy itself (writing-marketing writes it), app or SaaS screens (design-product-ui), or building the code (frontend-craft).

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Intent: Write one sentence per page type: who arrives, from where, what they need, what the page must make them do.

    line 20
  2. Distinctive:

    Validation: Check in a real browser: Core Web Vitals, keyboard, screen reader, reduced motion, 320 px width.

    line 36
  3. Distinctive:

    Never lazy-load the LCP image (verified) [SRC-WEB-014].

    line 52
  4. Distinctive:

    Every motion has a job and a reduced-motion variant.

    line 77

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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

Web Experiences

Design a public website from what the visitor came to do and what the page must prove, then make it fast, readable, and accessible before it is memorable. A site that wins on looks and loses on load time or keyboard access fails the people it was built for. Evidence levels, routing, and the quality model live in design-core; this skill does not repeat them.

Golden path

Order matters because each step constrains the next.

  1. Intent. Write one sentence per page type: who arrives, from where, what they need, what the page must make them do. Composition without it optimizes looks.
  2. Information architecture. List content and relationships before any wireframe; navigation is the on-screen expression of IA, not the IA itself (supported) [SRC-PROD-010]. Label menus with clear, familiar words (supported) [SRC-PROD-011].
  3. Composition. Derive the visual direction from the real content and brand, not from a template, because templated sameness is the first thing a returning visitor stops seeing (hypothesis) [SRC-WEB-029].
  4. Motion. Give each motion a job and a reduced-motion variant per the core motion guide, design-core/references/motion.md, before choosing any library, because motion chosen first drives the JS budget.
  5. Performance. Budget LCP, INP, and CLS at the 75th percentile before the hero is final (verified) [SRC-WEB-012], because the hero image and the motion library are what usually break them.
  6. Validation. Check in a real browser: Core Web Vitals, keyboard, screen reader, reduced motion, 320 px width.

Decision rules

  • Copy is owned by writing-marketing; leave placeholders with the job each text must do ("proof of claim X") rather than writing filler that later survives into production.
  • An award is a jury's subjective score - usability is 30% of it - not tested usability, accessibility, or performance (verified rubric) [SRC-WEB-024]. Copy a pattern from an award winner only after a real-browser check.
  • Content must exist in the initial HTML with one h1; a JS-only shell hides the page from readers and crawlers until the bundle runs (context-dependent, scan-based) [SRC-WEB-027].
  • Autoplaying motion over 5 s gets a pause control (verified) [SRC-WEB-003].
  • Never lazy-load the LCP image (verified) [SRC-WEB-014].
  • Urgency animation and fake scarcity are a trust breach, not a style (supported) [SRC-WEB-020].

Load when

  • references/web-golden-path.md - always, for a new page or site; it has the per-site-type path (marketing, landing, editorial, e-commerce, docs) and the worked example.
  • references/visual-craft.md - when choosing typography, grid, density, imagery, responsive behavior, or a visual direction, or when a draft looks like every other landing page.
  • references/awards-and-inspiration.md - when the brief cites Awwwards, an award site, or "make it feel like X", or when reviewing an expressive design.
  • references/performance.md - when setting budgets, choosing images or fonts, adding a motion or 3D library, or when a page feels slow.
  • Motion: design-core/references/motion.md for every animated surface.

Gates

Before done:

  • Intent sentence per page type and an IA list exist.
  • Every motion has a job and a reduced-motion variant; no autoplay over 5 s without pause [SRC-WEB-003].
  • Performance budget stated: LCP at most 2.5 s, INP at most 200 ms, CLS at most 0.1 at p75, mobile and desktop separately [SRC-WEB-012].
  • WCAG 2.2 AA checked per the core [SRC-A11Y-001].

Automated: Lighthouse or a lab trace for LCP, CLS, and a long-task view; axe or equivalent; a reduced-motion emulation pass. Manual: field CWV after launch (lab only approximates), screen-reader walk-through, whether the visual direction fits the brand, whether each motion earns its job.

Done when

The gates pass, each material recommendation carries an evidence level and a source or is marked hypothesis, what could not be measured is written as "not measured", and the copy brief is handed to writing-marketing with the job of each text slot.

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.