agentleFS
MartinOliveroMartinOlivero/saas-builder/skills/ui-design/SKILL.md

ui-design

Skill4 starsChanged 32 days agoMITView on GitHub

This skill should be used when making visual design decisions, creating components, defining styles, colors, typography, spacing, or building any interface — including dashboards, admin panels, and web apps. Trigger phrases include "design the UI", "build a component", "pick colors", "create a design system", "make it look good", "style this", "dashboard layout", "make a dark mode". It generates a design system before writing code and avoids generic AI aesthetics.

59 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use harness-design, graphic-design and 9 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • VictoriakaeySkill

    harness-design

    Use when designing or improving the action space, observation format, tool boundaries, and evaluation signals of an LLM agent. Covers the five layers of harness design: what tools the agent has, what it sees after acting, how each tool is scoped, how behavior is evaluated, and how to iterate when the agent misbehaves. Prevents the most common harness failures — agents that pick wrong tools, ignore critical information, or loop without progress.

    Also in 3 projects that use this

  • travisjneumanSkill

    graphic-design

    Professional graphic design principles for digital and print media. Use when creating visual designs, choosing color palettes, typography, layouts, or providing design feedback.

    Also in 3 projects that use this

  • soilmassSkill

    typography

    Typography and kinetic text animations including character reveals, typewriter effects, scramble/decode, wave motion, and animated gradients. Use when creating text reveals, headline animations, number counters, or any text-based motion effects.

    Also in 3 projects that use this

  • MicrosoftSkill

    typescript

    TypeScript best practices for this monorepo. Use this skill when: (1) writing new TypeScript code, (2) reviewing TypeScript code, (3) working with API responses, fetch calls, or JSON parsing, (4) adding or modifying types and interfaces.

    Also in 7 projects that use this

  • VercelSkill

    react

    React renderer for json-render that turns JSON specs into React components. Use when working with @json-render/react, building React UIs from JSON, creating component catalogs, or rendering AI-generated specs.

    Also in 4 projects that use this

  • 3stoneBrotherSkill

    code-audit

    Professional code security audit skill covering 55+ vulnerability types. Enhanced with WooYun 88,636 real-world vulnerability cases (2010-2016). This skill should be used when performing security audits, vulnerability scanning, penetration testing preparation, or code review for security issues. Supports 9 languages: Java, Python, Go, PHP, JavaScript/Node.js, C/C++, .NET/C#, Ruby, Rust. Includes 143 mandatory detection items across all languages with language-specific checklists. Covers SQL injection, XSS, RCE, deserialization, SSRF, JNDI injection, JDBC protocol injection, authentication bypass, business logic flaws, race conditions, and modern security domains (LLM, Serverless, Android). WooYun integration adds: statistical-driven parameter priority, bypass techniques library, logic vulnerability patterns, and real-case references. v1.0: Initial public release with Docker deployment verification framework.

    Also in 3 projects that use this

The file

Show its rules and text

We split this skill into the rules it gives an agent

14 rules. 0 are common, 8 are distinctive, and 6 are very specific to this project.

  • 0

    Common

    Most of the 50 edited versions we read keep these.

  • 8

    Distinctive

    This author's own choices. Worth a look.

  • 6

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Always prefer the originals when present.

    line 20
  2. Very specific:

    If ui-ux-pro-max is installed, let it produce the system, then optionally refine with taste-skill.

    line 31
  3. Distinctive:

    If neither is installed, apply the embedded fallback below.

    line 32
  4. Very specific:

    Default stack: React + Vite + Tailwind + shadcn/ui.

    line 44
  5. Very specific:

    Dashboard mode: when building a dashboard, admin, or backoffice, bias VISUAL_DENSITY high (7-9), keep DESIGN_VARIANCE moderate (data …

    line 54
  6. Distinctive:

    Magic colors: no raw hex values scattered in markup.

    line 61
  7. Distinctive:

    Ignored states: never deliver a UI without designing the empty, loading, and error states.

    line 62
  8. Distinctive:

    A screen that lies: never render "could not find out" as zero, empty or off.

    line 63
  9. Distinctive:

    Every piece of data on screen is in one of three conditions, and each needs its own rendering:

    line 67
  10. Distinctive:

    Show the time of the last successful sync: instead of asserting the data is current.

    line 79
  11. Distinctive:

    Don't scatter raw hex. Use a three-tier token architecture so rebrands and dark mode are one change, not a hundred:

    line 87
  12. Very specific:

    Prefer container queries (@container) over media queries for reusable components, so a card adapts to its slot, not the viewport.

    line 98
  13. Very specific:

    Use intrinsic layout primitives (Grid auto-fit/minmax, Flexbox) over manual breakpoints.

    line 102
  14. Very specific:

    When building landing or marketing pages, the standard component set is: hero, navbar, features, pricing, testimonials, CTA, footer.

    line 112

“Common” is measured: kept by at least 6 in 10 of the 50 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

UI Design

This skill produces distinctive, production-grade interfaces. It orchestrates a UI/UX-pro approach with deliberate taste: decide the design system first, then write code that expresses it.

Analogy: an architect does not start laying bricks. They draw the blueprint — materials, proportions, palette — and only then build. The design system is the blueprint.

Dependencies

This skill works standalone, but it is significantly stronger when the original design skills are installed. Before generating a design system, check what is available and delegate in this order:

  1. ui-ux-pro-max available → invoke it to generate the full design system (palettes, styles, typography). It has the richest design database.
  2. taste-skill available → apply it as a style layer on top of the system from step 1 (or on top of the embedded fallback).
  3. Neither installed → use the embedded design principles in this skill as the fallback. They are kept below precisely for this case.

Always prefer the originals when present; fall back gracefully and silently when they are not. Never block on a missing dependency.

Workflow

Step 1 — Identify the product type

The product type drives the entire visual language. Classify first: fintech, edtech, e-commerce, SaaS, health, dev tool, social, creative, internal/dashboard, etc. A health app and a crypto exchange should never look the same.

Step 2 — Generate the design system BEFORE writing code

First, resolve the design system per the Dependencies order above:

  • If ui-ux-pro-max is installed, let it produce the system, then optionally refine with taste-skill.
  • If neither is installed, apply the embedded fallback below.

Embedded fallback — define and state these explicitly:

  • Visual style — pick one with intent: glassmorphism, minimalism, brutalism, bento grid, neo-retro, editorial, etc. Justify it against the product type.
  • Color palette — primary, secondary, neutrals, and semantic colors (success / warning / error / info). Defined as tokens, not ad-hoc hex.
  • Typography — a heading font, a body font, and a mono font (for code/numbers). Avoid defaulting to Inter for everything.
  • Spacing and border radius — a consistent scale (e.g. 4/8/12/16/24) and a radius decision (sharp, soft, or pill).
  • Dark mode — decide yes or no up front; if yes, define the dark tokens now, not later.

Step 3 — Apply to the user's stack

Default stack: React + Vite + Tailwind + shadcn/ui. Supported alternatives: Next.js, Vue, Svelte, plain HTML + Tailwind. Ask which one if unknown; otherwise use the default and say so.

Step 4 — Tune the three design parameters

Expose and set these 1-10 dials, and let the user adjust:

  • DESIGN_VARIANCE (1-10) — from centered, safe layouts (1) to asymmetric, expressive compositions (10).
  • MOTION_INTENSITY (1-10) — from simple hover states (1) to scroll-driven and orchestrated animations (10).
  • VISUAL_DENSITY (1-10) — from airy, spacious layouts (1) to compact, information-dense dashboards (10).

Dashboard mode: when building a dashboard, admin, or backoffice, bias VISUAL_DENSITY high (7-9), keep DESIGN_VARIANCE moderate (data clarity over expression), and prioritize legible tables, filters, empty/loading/error states, and scannable hierarchy.

Step 5 — Respect the forbidden anti-patterns

Never ship these:

  • The AI default look: Inter + purple gradient + heavy rounded cards. It screams "generated." Avoid it.
  • Magic colors: no raw hex values scattered in markup. Every color comes from the system tokens.
  • Ignored states: never deliver a UI without designing the empty, loading, and error states. They are part of the work, not an afterthought.
  • A screen that lies: never render "could not find out" as zero, empty or off. See Honest states below.

Honest states — three answers, not two

Every piece of data on screen is in one of three conditions, and each needs its own rendering:

ConditionExampleRender
Known: yes3 agents onlineThe value
Known: no0 agents onlineThe empty state
UnknownThe request failed or timed out"Couldn't check" + retry — never the empty state

The third is the one that gets dropped: a failed request becomes an empty list, and the interface states something false with full confidence — "0 of 5 online", "no team members", "nothing was changed". In one production app this pattern was fixed eight separate times before it became a rule.

Apply it to writes too:

  • Confirm only what was confirmed. "Invitation sent" only if the provider said so.
  • Show the time of the last successful sync instead of asserting the data is current.
  • After a failed write to an external system, say "couldn't confirm", not "nothing changed" — the user will retry.
  • Name the specific error for the specific case. "Link unavailable" for a 404 only; anything else is a connection problem.

Some defects only exist in the final artifact, so check the artifact: print styles that produce a blank page, a social image rendered with a fallback font, a font subset that silently falls back. A passing build says nothing about any of them.

Design tokens — structure them in three tiers

Don't scatter raw hex. Use a three-tier token architecture so rebrands and dark mode are one change, not a hundred:

  1. Primitive — raw values (blue-500: #3b82f6).
  2. Semantic — role aliases (color-primary, color-bg-surface, color-danger). Components reference only these, never primitives.
  3. Component — optional per-component (button-bg).

In this stack, expose tokens as CSS custom properties wired into Tailwind v4's @theme block (--color-primary, --spacing-*, --radius-*) so utilities and tokens share one source. Theme (light/dark, multi-brand) by swapping CSS-variable values under [data-theme] / .dark — component classes stay identical. Name tokens by role, not appearance (color-danger, not color-red). Cover color, spacing, typography, radius, shadow, z-index, and motion duration/easing. Only reach for Style Dictionary if you must emit tokens to multiple platforms (web + iOS/Android/email); web-only, the Tailwind @theme path is enough.

Responsive design — mobile-first, always

  • Author base styles for small screens; layer up with min-width / container queries. Never desktop-down.
  • Prefer container queries (@container) over media queries for reusable components, so a card adapts to its slot, not the viewport.
  • Fluid type/space with clamp(min, preferred-vw, max) instead of breakpoint-stepped sizes.
  • Touch targets ≥ 44×44px (Apple) / 48×48px (Material), ≥ 8px apart.
  • Handle notches with env(safe-area-inset-*) + viewport-fit=cover; pad sticky headers/footers.
  • Use intrinsic layout primitives (Grid auto-fit/minmax, Flexbox) over manual breakpoints.

(For deeper accessibility, performance, or installable/offline behavior, hand off to the accessibility, frontend-performance, and pwa skills.)

Output

Produce ready-to-use components, not throwaway prototypes — wired to the design tokens, with all states handled.

Key landing components

When building landing or marketing pages, the standard component set is: hero, navbar, features, pricing, testimonials, CTA, footer. (For full landing structure and copy, the landing-page skill drives the flow and calls this skill for the visual system.)

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.