What it is
ai-design-skills is a small collection of Markdown design-skill files for coding agents. At the research date it contained exactly one skill, landing-page-design. That skill combines intake questions, page structure, copy rules, and a strict visual system covering fonts, spacing, radius, color, and motion.
It is not an extractor or a token file. It is a prose playbook meant to steer an agent’s choices while it designs or builds a landing page. The README says more skills are planned. It points to a separate redesign skill repository, which already existed, for changing an existing site.
First principles
A skill file can change an agent’s output only through the text it adds to the agent’s context. Three constraints follow.
- It has to be loaded. An agent decides whether to load a skill from its frontmatter
description. A broad description loads the skill more often, including on tasks where it does not fit. - It has to beat the model’s defaults. Exact values, such as a spacing scale or an easing curve, change output. Vague advice, such as asking for a clean layout, mostly restates what the model would do anyway.
- It costs context every time it loads. A single long file is read in full on each use. Splitting a skill into a short entry file plus resources read on demand keeps that cost down.
This skill takes a clear position on all three: a very broad trigger, very specific values, and one monolithic file. The sections below show where each choice helps and where it hurts.
Who is behind it
The repository is owned by Elaya, through the elayadesign GitHub handle. The license says “Copyright (c) 2026 Elaya.” The repository gave no company, team, or founder information beyond that handle and copyright. Further identity claims are UNVERIFIED.
As of September 2026, the MIT-licensed repository had 2,302 stars, 159 forks, and one open issue. It was created and last pushed on 2026-07-29. Its full commit history had three commits, all from the day it was created. The repository is highly visible but has been inactive since launch.
Format and how an agent consumes it
Each skill is a folder with a single SKILL.md. The frontmatter is deliberately broad, so that an agent loads it whenever a task touches a landing page:
---
name: landing-page-design
description: "Complete system for building high converting landing pages: intake questions, page structure, layout selection, conversion copywriting, SEO, plus strict visual rules for typography, spacing, corner radius, backgrounds, hero layout, icons, and motion. Use this skill whenever building, editing, styling, reviewing, or writing copy for ANY landing page, marketing site, web UI, page section, component, or prototype, even when the user does not mention design, conversion, tokens, or a design system. Always consult it before choosing page sections, headlines, CTAs, fonts, font sizes, spacing values, border radii, background colors, icons, or transitions."
---The body carries concrete design instructions, not machine-parsed tokens. For example:
| Token | Value |
|---|---|
| Spacing-0 | 0 |
| Spacing-25 | 2px |
| Spacing-50 | 4px |
| Spacing-75 | 8px |
| Spacing-100 | 12px |
| Spacing-200 | 16px |It also prescribes the nested-radius formula and a motion primitive:
inner radius = outer radius − gap
transition-all duration-700 ease-[cubic-bezier(0.32,0.72,0,1)]The README explains how to copy the file into each tool. A Claude Code installation example is:
mkdir -p .claude/skills
git clone https://github.com/elayadesign/ai-design-skills.git /tmp/ai-design-skills
cp -r /tmp/ai-design-skills/skills/<skill-name> .claude/skills/Cursor can save the same file into .cursor/rules. The skill is also presented as adaptable for Codex, Windsurf, Cline, and Claude Projects. It is not a fully bundled Claude Skills package: it has no scripts, no resources, and no progressive disclosure structure.
Pricing and access
The project is free and MIT licensed. Its README says users can use it, fork it, ship it, and sell what they build, with no attribution required. There is no paid tier, hosted service, or command-line interface (CLI).
What works
The specificity is the point. Exact spacing values, colors, curves, typography restrictions, content rules, and a mandatory output order constrain generic agent output better than vague design advice. The skill also covers strategy and visuals together, so an agent is not asked to design a hero section without considering page structure or copy.
One file serving several tools is a low-friction way to distribute. The dense frontmatter is a deliberate attempt to make automatic loading likely.
What does not
The collection has one skill despite describing itself as growing. It has had no commits since its creation day. Whether the maintainer responds to issues is UNVERIFIED. The single monolithic file does not use more advanced skill packaging. Its strong font bans can also conflict with an existing brand unless an override rule is supplied.
Its advice is prescriptive, not a portable token system. A team cannot assume the skill reflects its actual product.
What we take from it
Velofy could propose a single source file with clear install paths across agents, plus specific tokens and motion rules that change output in practice. It could propose versioning and a real collection of focused skills, not a promise of expansion.
Velofy has not published a design skill, package, or Model Context Protocol (MCP) server. A future proposal should keep skill packaging open, state an override policy, and distinguish a style guide from evidence extracted from a product.