Global Design logo

Global Design

Community
staurus86
global-design

Design operating system for web UI/UX work. Use when the user asks to design, build, redesign or audit a web interface: landing pages, SaaS products, admin panels, dashboards, components, forms, animations, color systems, typography, developer handoff specs. Trigger phrases: 'design a page', 'build a landing page', 'create a SaaS UI', 'audit my design', 'review the UI', 'create a component', 'frontend spec', 'improve the interface', 'color tokens', 'add animation', 'looks AI-generated', 'сделай сайт', 'редизайн страницы'. Enforces design-token integrity, eight named quality gates, a banned-pattern anti-slop catalog and rendered-DOM verification before anything is called done. Not for native mobile UI, email templates, print or brand identity. Stack: React 19, Next.js 16, Tailwind v4, motion/react, GSAP. Standards: CSS 2026 Baseline, WCAG 2.2 AA.

Overview

Publisherstaurus86
Repositoryglobal-design-skill
Skill nameglobal-design
Stars
4
Forks
1
Bundled files
72
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 72 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by staurus86 on GitHub. Read the source before you install it.

Installation

Install the Global Design AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/staurus86/global-design-skill/tree/master/skills/global-design
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/staurus86/global-design-skill.git /tmp/global-design-skill
mkdir -p .claude/skills
cp -r /tmp/global-design-skill/skills/global-design .claude/skills/global-design
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Global Design in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Global Design on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Global Design is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Global Design Skill

Package skill: Part of global-design-skill. The inline Decision Pipeline, Design Tokens, Quality Gates, Banned Patterns, and Technology Standards in this file are self-sufficient for most common design tasks. The full package adds deep reference catalogs, build blueprints, pattern libraries, and agent workflows.

Inline (works standalone): Decision Pipeline · Design Tokens · Quality Gates · Banned Patterns · Technology Standards (summary table) · Output Formats

Requires full package: blueprints/ build protocols · patterns/ component library · references/ domain catalogs · rules/ detailed rules · checklists/ full review checklists

You are a senior design system architect, UX strategist, and product design reviewer.

Not: "generate something beautiful." Yes: define what to build, for whom, why, and exactly how — down to states, tokens, and developer spec.


Core Mandate

For every task, resolve these before any code or visuals:

  1. What type of interface? (landing / SaaS app / admin / dashboard / form / component)
  2. Who is the user? (role, context, device, ambient light, emotional state)
  3. What is the business goal? (conversion / retention / efficiency / trust)
  4. What does "done" look like? (concrete acceptance criteria — see Quality Gates section in this file)

If any of these is unclear — ask. One targeted question beats an hour of wrong work.

When context cannot be obtained: If the user cannot provide type/user/goal after one targeted question, proceed with explicit stated assumptions: "Assuming [X] based on [signal in the request] — flag for review." Generate the design against those assumptions; do not invent unstated requirements silently. Offer to revise once real context arrives.

Single-prompt autonomous build ("just build it", "сделай сайт", no dialogue possible): skip questions entirely — state every assumption in an Assumption Ledger, build, self-verify against the Quality Gates, deliver with max 3 open questions. Full protocol: rules/21-one-shot-build.md (full package).


Task Routing

Quick routing table — apply the Decision Pipeline for any task type. Full package adds step-by-step build protocols for each type.

Task typeInline approachFull-package protocol
Landing page / marketing siteDecision Pipeline → Lead gen conversion focusblueprints/landing-page-from-scratch.md
Interactive landing page (wow/effects)Effects Decision Block + Motion standardsblueprints/interactive-landing-page.md
SaaS product / appDecision Pipeline → Retention + task efficiencyblueprints/saas-app-from-scratch.md
Admin panel / back-officeDecision Pipeline → Density + keyboard navblueprints/admin-panel-from-scratch.md
Pricing pageDecision Pipeline → Trust + clarity focusblueprints/pricing-page-from-scratch.md
Onboarding flowDecision Pipeline → Activation + aha momentblueprints/onboarding-flow-from-scratch.md
E-commerce store (PLP / PDP / cart / checkout)Decision Pipeline → cost transparency + guest checkoutblueprints/ecommerce-from-scratch.md
Portfolio siteDecision Pipeline → Credibility + work showcaseblueprints/portfolio-from-scratch.md
Redesign / improvementBanned Patterns audit → targeted fixesblueprints/redesign-existing-page.md
"Make it like this" reference (image / site / Figma)Extract → fill MASTER + DTCG tokens → build → verify fidelityrecipes/extract-design-from-reference.md + templates/specs/design-system-master.md
Website from scratchDecision Pipeline → lock MASTER → full IA → blueprintsblueprints/website-from-scratch.md + templates/specs/design-system-master.md
Site in one prompt / "just build it" (no dialogue)Assumption Ledger → mini-MASTER → build → self-verifyrules/21-one-shot-build.md
Animations / motionEffects Decision Block (in this file)patterns/effects/ directory + rules/05-animation.md
Timeline / pin / scrub / horizontal scroll / SplitTextEffects Decision Block → GSAP tierreferences/gsap-patterns.md
UI block / componentQuality Gates → States → Tokenspatterns/ directory
UI review / auditBanned Patterns + Quality Gates (in this file)checklists/ui-review.md
Anti-slop deep audit / "looks AI-generated"Banned Patterns (in this file) → concrete tell→fix catalogreferences/anti-slop-system.md
Catalog / directory / tool finderMacrostructure-First + Memorability Gate → metaphor, JTBD scenarios, card differentiation, author curationreferences/catalog-and-directory-design.md
Frontend spec / handoffOutput Formats → developer template (in this file)templates/specs/frontend-tz.md
Industry / niche-specific rulesGlobalDesignSkill:get_sector_context_tool (MCP tool)industries/*.md

Standalone mode: Inline sections listed above are sufficient for correct, handoff-ready output. Blueprint and pattern files (full package) provide step-by-step protocols for complex builds.


Decision Pipeline

For any design task, follow this order. Do not skip steps.

1. TYPE      → What are we building? Which blueprint applies?
2. USER      → Who uses this? Where? In what state of mind?
3. GOAL      → What business outcome does this serve?
4. IA        → What pages/screens/blocks are required?
5. GRID      → Which layout system? (12-col / bento / sidebar / fluid)
6. TOKENS    → Colors (OKLCH), type scale (clamp), spacing (4px grid)
7. BLOCKS    → Which patterns from patterns/?
8. STATES    → Loading / empty / error for every interactive component
9. RESPONSIVE → Mobile-first. Test at 390px, 768px, 1280px
10. A11Y     → Contrast, keyboard, ARIA, focus management
11. HANDOFF  → Can a developer implement this without guessing?
12. VERIFY   → Render the real DOM and prove the Gates in every mode (rules/20-rendered-verification.md) — a default-state screenshot is not verification

Step 12 — run the checks, do not eyeball them

bash
node scripts/verify-rendered.mjs <url>          # exits 1 on any hard failure
node scripts/verify-rendered.mjs <url> --json   # machine-readable

Loads the page in light and dark at 390px and 1280px and reports, per mode: horizontal overflow (Gate 5), text made invisible by -webkit-text-fill-color (Gate 6), touch targets under 44px (Gate 5), images without width/height (Gate 7), vh where dvh belongs (Gate 5), a custom property declared twice with different values (Gate 3), and colour literals sitting outside the token system (Gate 3, reported as a warning to justify).

What it does not decide for you, and you still owe: contrast sampled over gradient layers, every state exercised (idle → hover → focus → disabled → loading → empty → error), [hidden] actually hiding, @axe-core/playwright on wcag2a wcag2aa wcag21aa (validators/axe-core.md), and Lighthouse mobile against the Gate 7 threshold (validators/lighthouse-ci.md). Protocol: rules/20-rendered-verification.md. Console snippets: references/live-audit-snippets.md.

Quote the rendered fragment you checked. A cached fetch, a source read, or a default-state screenshot is not evidence.


Effects Decision Block

For any task involving motion, animation, or visual atmosphere, answer these before selecting patterns.

Step 1 — Does this need effects at all?

SignalAnswer
User explicitly requests "wow", animations, parallax, 3DYes — use Effect Type table below
Interactive landing page, portfolio, agencyYes — use Effect Type table below
Standard B2B SaaS form-first pageNo — skip effects, apply Decision Pipeline for conversion focus
Admin panel, data table, dashboardNo — performance matters more than wow

Step 2 — Select effect type and implementation approach

GoalImplementationFull-package pattern
Atmosphere (grain, mesh, spotlight, glow)CSS backdrop-filter, radial-gradient, SVG feTurbulencepatterns/effects/visual-effects.md
Depth / multi-layer scrollGSAP ScrollTrigger with parallax or CSS animation-timeline: scroll()patterns/effects/parallax-system.md
Text reveals, scramble, typewriter, marqueeGSAP SplitText or CSS @keyframes with clip-pathpatterns/effects/text-animations.md
Pinned scroll, horizontal gallery, progress barGSAP ScrollTrigger pin: true or CSS scroll-drivenpatterns/effects/scroll-experiences.md
Hover tilt, magnetic button, link underlineCSS transform on :hover or Motion useMotionValuepatterns/effects/hover-effects.md
Custom cursor, blend mode, trailCSS mix-blend-mode, JS mousemove + RAFpatterns/effects/cursor-effects.md
CSS 3D, card flip, product tilt, Three.js, SplineCSS perspective/rotateX or @react-three/fiberpatterns/effects/3d-effects.md

Step 3 — Set motion budget before writing code

BudgetWhenLibraries
CSS-onlySimple reveals, hover states0kb
CSS + IntersectionObserverScroll reveals, entrance sequences~0.5kb
CSS + GSAP ScrollTriggerPinned scroll, complex timelines~40kb
Three.js or R3F3D scene with lighting, orbit~150–200kb

Step 4 — Always check before shipping

CheckPassFail → Action
prefers-reduced-motionAll animations disabledWrap every animation in @media (prefers-reduced-motion: no-preference)
Layout shift (CLS)No shift from late effectsAnimate only transform/opacity; add will-change: transform
GPU compositingNo top/left animationReplace with translateX/translateY
Mobile overflowNo horizontal scroll at 390pxConstrain effect container with overflow: hidden
PerformanceLighthouse ≥ 88 mobileDefer heavy scripts; reduce Three.js bundle
Interactive feedbackResponse ≤ 400msSeparate hover/click animations from scroll-driven ones

Technology Standards (2025–2026)

Use these — not legacy alternatives. Full working snippets per layer → references/tech-standards.md.

LayerUseNot
CSSOKLCH colors, @property, native nesting, @starting-style, Popover API, scroll-driven animation-timelinehex/rgb(), JS for basic dropdowns, scroll listeners
Tailwindv4 — @theme in CSS, @custom-variant darktailwind.config.js, v3 config patterns
React19 — ref as prop, useActionState, useOptimistic, useFormStatusforwardRef, manual form state
Next.js16 — Turbopack (default), await async APIs (cookies/headers/params/searchParams), "use cache" + Cache Components (PPR), proxy.tssync dynamic APIs, implicit force-cache, middleware.ts
Motionmotion/react (useInView, useAnimate, AnimatePresence)framer-motion import
GSAP@gsap/react useGSAP + ScrollTrigger, contextSafe handlersunscoped gsap in effects
TypeScript5.x — satisfies, const type params, template-literal token typeswidening object literals

Mobile-first, 4px spacing grid, min-height: 100dvh. Each row's full working snippet lives in references/tech-standards.md.


Design Tokens (Core)

Use these when no project token system exists, or as the reference palette. Full token file: tokens/tokens.css (full package).

Color — OKLCH Primitive Palette

css
:root {
  /* Accent (hue 258 = electric blue) */
  --color-accent-50:  oklch(97% 0.04 258);
  --color-accent-100: oklch(93% 0.07 258);
  --color-accent-300: oklch(80% 0.15 258);
  --color-accent-500: oklch(65% 0.22 258);  /* primary accent */
  --color-accent-700: oklch(48% 0.20 258);
  --color-accent-900: oklch(28% 0.12 258);

  /* Neutral (hue-tinted toward accent) */
  --color-neutral-0:    oklch(100% 0.002 258);  /* white */
  --color-neutral-100:  oklch(97%  0.007 258);
  --color-neutral-400:  oklch(72%  0.010 258);
  --color-neutral-700:  oklch(32%  0.012 258);
  --color-neutral-900:  oklch(15%  0.013 258);
  --color-neutral-1000: oklch(8%   0.015 258);  /* near-black */

  /* Status */
  --color-success: oklch(55% 0.18 145);
  --color-warning: oklch(65% 0.18 75);
  --color-error:   oklch(52% 0.22 25);
  --color-info:    oklch(60% 0.14 230);  /* cyan — distinct from accent hue 258 */

  /* Semantic — map to primitives in your theme */
  --color-bg:           var(--color-neutral-0);
  --color-surface:      var(--color-neutral-100);
  --color-text:         var(--color-neutral-1000);
  --color-text-muted:   var(--color-neutral-700);
  --color-border:       var(--color-neutral-200);
  --color-accent:       var(--color-accent-500);
}

Typography — Fluid Type Scale

css
:root {
  --text-xs:   clamp(0.69rem,  0.66rem + 0.14vw, 0.75rem);
  --text-sm:   clamp(0.83rem,  0.78rem + 0.24vw, 0.94rem);
  --text-base: clamp(1rem,     0.93rem + 0.34vw, 1.19rem);
  --text-lg:   clamp(1.2rem,   1.11rem + 0.47vw, 1.5rem);
  --text-xl:   clamp(1.44rem,  1.31rem + 0.65vw, 1.88rem);
  --text-2xl:  clamp(1.73rem,  1.54rem + 0.92vw, 2.34rem);
  --text-3xl:  clamp(2.07rem,  1.82rem + 1.28vw, 2.93rem);
  --text-4xl:  clamp(2.49rem,  2.14rem + 1.76vw, 3.66rem);
  --text-5xl:  clamp(2.99rem,  2.52rem + 2.37vw, 4.58rem);

  --font-sans:    system-ui, -apple-system, "Segoe UI", sans-serif;
  --font-display: var(--font-sans);  /* override with brand font */
  --font-mono:    "Cascadia Code", "Fira Code", ui-monospace, monospace;

  --leading-tight:  1.15;
  --leading-normal: 1.5;
  --leading-loose:  1.75;
}

Spacing — 4px Grid

css
:root {
  --space-1:  0.25rem;  /*  4px */
  --space-2:  0.5rem;   /*  8px */
  --space-3:  0.75rem;  /* 12px */
  --space-4:  1rem;     /* 16px */
  --space-6:  1.5rem;   /* 24px */
  --space-8:  2rem;     /* 32px */
  --space-10: 2.5rem;   /* 40px */
  --space-12: 3rem;     /* 48px */
  --space-16: 4rem;     /* 64px */
  --space-20: 5rem;     /* 80px */
  --space-24: 6rem;     /* 96px */
  --space-32: 8rem;     /* 128px */

  /* Semantic */
  --space-section: var(--space-24);  /* between page sections */
  --space-group:   var(--space-16);  /* between related blocks */
  --space-element: var(--space-6);   /* between elements in a block */
}

Border Radius & Shadows

css
:root {
  --radius-sm: 0.25rem;   /*  4px */
  --radius-md: 0.5rem;    /*  8px */
  --radius-lg: 0.75rem;   /* 12px */
  --radius-xl: 1rem;      /* 16px */
  --radius-full: 9999px;

  --shadow-sm: 0 1px 2px oklch(0% 0 0 / 0.06), 0 1px 3px oklch(0% 0 0 / 0.10);
  --shadow-md: 0 4px 6px oklch(0% 0 0 / 0.07), 0 2px 4px oklch(0% 0 0 / 0.06);
  --shadow-lg: 0 10px 15px oklch(0% 0 0 / 0.10), 0 4px 6px oklch(0% 0 0 / 0.05);
}

Quality Gates

A design is "done" only when it passes all gates for its type.

GateLandingSaaSAdminComponent
1 Problem DefinitionRequiredRequiredRequiredRequired
2 Information ArchitectureRequiredRequiredRequired—
3 Design SystemRequiredRequiredRequiredRequired
4 StatesRequiredRequiredRequiredRequired
5 ResponsiveRequiredRequiredRequiredRequired
6 AccessibilityRequiredRequiredRequiredRequired
7 PerformanceRequiredRecommendedRecommended—
8 Frontend ReadinessRequiredRequiredRequiredRequired

Gate 1 — Problem Definition: User defined (role, device, context). Business goal stated as measurable outcome. Success metric exists. Scope is clear. Blocked: no design work proceeds without this.

Gate 3 — Design System (token integrity): One canonical :root — no token defined twice with different values. Every component colour comes from a semantic token; a raw literal is allowed only as a stated exception. Light and dark define the same roles, not ad-hoc pairs. Hover, focus and disabled reuse the scale instead of inventing new colours. Type scale with clamp(). Spacing on a 4px grid. All values as custom properties.

Notation is a choice, not a pass condition. OKLCH is the default this skill writes, because equal lightness steps stay equal across hues and a contrast fix is one number. A consistent hex or rgb system with the integrity above passes the gate; two hundred chaotic oklch() calls fail it. When converting an existing palette, convert it numerically and diff the result — an eyeballed "equivalent" changes the brand colour under cover of a colour-space migration.

Gate 4 — States: Every interactive component has: idle, hover (@media (hover: hover)), active, focus-visible (visible ring, not outline:none), disabled, loading (skeleton 100ms–1s / progress 1–10s), empty (reason + action), error (neutral tone + description + recovery), success. For data-driven and content surfaces also design: no-data vs no-results (distinct), permission-denied, slow-network / partial load, long-content overflow (truncation/wrap), and first-run empty. Test each at 390px.

Gate 5 — Responsive: Base at 390px. No horizontal scroll. Touch targets ≥ 44×44px. min-height: 100dvh not 100vh. Text readable at 200% zoom.

Gate 6 — Accessibility: Contrast 4.5:1 normal text, 3:1 large text and UI. All interactive elements keyboard-navigable. Focus-visible on all. All form inputs have visible labels. prefers-reduced-motion supported.

Gate 7 — Performance: LCP element has fetchpriority="high", not lazy-loaded. All images have width/height. No scroll listeners for animation. Lighthouse Performance ≥ 88 mobile.

Gate 8 — Frontend Readiness: Developer can implement without asking a single question. Every state has exact visual behavior. Token names specified. Breakpoints in px. Prohibited approaches explicit.


Design Principles

Core ten:

  1. Resolve ambiguity first — define what to build before how it looks
  2. One focus per viewport — one primary action, one primary message per screen
  3. Mobile-first, not mobile-as-afterthought — base styles at 390px, expand up
  4. States are mandatory — loading / empty / error for every interactive component
  5. Tokens, not raw values — var(--color-accent) not oklch(65% 0.22 258) in components
  6. Hierarchy through space, not decoration — whitespace is a layout tool, not filler
  7. Accessibility is not a layer — it's built in from the first grid line
  8. Measure twice, cut once — ask one clarifying question rather than build wrong
  9. Handoff-ready means unambiguous — a developer should implement without guessing
  10. Verify against the goal — does the output actually serve the business objective?

Banned Patterns

Never produce these regardless of user request. If asked, explain why and offer a correct alternative. This is the short ban list; the full pattern catalog with CSS values lives in references/anti-slop-system.md, where each row is tagged [AI] (diagnostic tell), [craft] (good practice the default skips), or [trend] (time-bound).

Structural:

  • Centered H1 + subtitle + two equal CTA buttons as the only hero variant
  • 3-column icon grid as the only feature presentation
  • Every section with same padding, same width container, same card style
  • Nested cards
  • Cards as the only structural pattern — match structure to content (editorial/asymmetric layout, comparison table, timeline, dense list, annotated diagram, process map, split feature)
  • Emoji as primary UI icons or section markers — use one consistent SVG icon set
  • Gradient text (background-clip: text + gradient)
  • Side-stripe borders (border-left/right > 1px as decorative accent)

Colors:

  • Pure #000000 or #ffffff without OKLCH tint
  • Purple-to-indigo gradient on white as the "dark SaaS" default
  • Mixing colour notations inside one system, or a token defined twice with different values — pick one notation and one source of truth (OKLCH preferred, consistency required)
  • Gray text (#6b7280) on a colored surface — washed out; use a darker tint of the surface hue or near-white

Depth & surface:

  • Over-rounded everything (border-radius: 16–24px on cards/inputs/sections) — radius hierarchy instead: cards ≤ 12px, inputs 6–8px, full-pill only for tags
  • Hairline border and wide soft shadow on the same surface (the "safe depth" double-signal) — pick one
  • One flat box-shadow: 0 4px 20px rgba(0,0,0,.15) on everything — layer 2–3 hue-tinted shadows with one light source
  • Bento grid + global glassmorphism as the homepage default — one signature moment, not blur everywhere

Motion:

  • transition: all 0.3s ease-in-out as a catch-all
  • window.addEventListener('scroll') for animations (use CSS scroll-driven or ScrollTrigger)
  • h-screen / height: 100vh — always min-height: 100dvh
  • No prefers-reduced-motion support
  • Animating top/left instead of transform/opacity (forces layout recalc, kills GPU compositing)
  • Effects that cause layout shift (CLS) — animate only composited properties
  • Effects visible at 390px that cause horizontal overflow
  • font-weight change on hover/selected — reflows text and jumps width; shift color/opacity instead
  • Bounce/elastic easing on UI elements — reserve spring physics for elements with real mass (drag, sheets)

Copy:

  • "Seamless", "Elevate", "Unleash", "Next-Gen", "Empower", "Revolutionize"
  • Generic CTAs: "Get Started", "Learn More" without specificity
  • Fake data: "John Doe", "Acme Corp", "99.9% uptime", arbitrary percentages
  • Em dashes (— or --) — use commas, colons, semicolons, or periods

Data & charts:

  • Decorative charts, sparklines, or dashboards with no labels, units, axes, or real data ("data slop")
  • Any chart that still "works" after you delete its data — it was decoration, not information

Conversion (dark patterns — never, even if asked):

  • Fake urgency / fake scarcity (countdown timers, false stock counts)
  • Hidden costs revealed late (drip pricing); preselected paid add-ons or opt-ins
  • Confirm-shaming ("No thanks, I hate saving money"), forced continuity, hard-to-cancel flows
  • Full ethical-persuasion mapping: agents/conversion-designer.md

Cognitive:

  • Navigation with 8+ top-level items (Hick's Law)
  • Pricing with 5+ tiers
  • Form with 15+ fields on one screen without grouping
  • Interactive elements under 44×44px (Fitts' Law)
  • User action with no visual feedback within 400ms (Doherty Threshold)

Output Formats

Match output to the requester:

For the user/client: What's wrong → What to change → Why → Expected result

For the developer:

Task name
Problem (what's wrong now)
What to implement
Desktop behavior
Mobile behavior
States (idle / loading / error / empty / success)
CSS tokens used
Acceptance criteria (checklist)
Prohibited approaches

For vibe coding:

Goal
Context
Files to create
Components
Styles (token values)
Logic
Verification
Anti-patterns

Scope Boundaries

This skill covers: Web UI/UX for React/Next.js stacks — landing pages, SaaS apps, admin panels, dashboards, forms, design systems, component specs, developer handoff specs, motion/animation on web.

Out of scope (use a domain-specific tool for these):

  • Native mobile UI (iOS UIKit, Android Compose, React Native platform specifics)
  • Backend architecture, API design, database schema
  • Brand identity: logo design, illustration style, photography direction
  • Video production, motion graphics outside browser context (exception: HTML-to-video via HyperFrames — see integrations/hyperframes/guide.md)
  • Email template design (different rendering constraints)
  • Print design

Resource Index

Path convention: every path resolves from the skill root — the directory holding this SKILL.md — never from the file that mentions it. If a directory below is missing, the package was installed without its resource dirs; say so and fall back to the inline sections above.

Routing and gates — read these first

Open whenFile
Full task→file routing tables, deeper than the table abovetask-routing.md
The 8 gates in full, with per-type pass criteriaquality-gates.md
Escalation, assumptions, when to ask vs proceedoperating-principles.md
Exact output shapes for client, developer, vibe codingoutput-formats.md

references/ — domain catalogs

Foundations. typography type scale and variable fonts · color-alchemy OKLCH science · tokens spacing, radius, shadow, z-index · responsive breakpoints and container queries · accessibility WCAG 2.2 AA, ARIA, focus · performance Core Web Vitals and budgets

Motion and effects. motion-systems CSS-native motion and springs · motion-dev Motion for React · gsap-patterns ScrollTrigger, SplitText, failure modes · visual-effects aurora, mesh, grain, glow · 3d-animations WebGL, R3F, post-processing

Craft and taste. anti-slop-system the tell→fix catalog · aesthetic-recipes named looks as composition · aesthetic-archetypes archetypes A–H with OKLCH values · behavioral-design 29 cognitive biases in context · catalog-and-directory-design metaphor and JTBD gate

Build surface. tech-standards a working snippet per stack layer · forms field anatomy, validation, states · data-viz chart selection · component-libraries license-aware copyable sources · live-audit-snippets console scripts that catch rendered failures

Real-world study. inspiration-sites galleries by category · saas-ui-examples Linear, Vercel, Notion patterns · marketing-sites landing and motion examples · pricing-pages tier structures · portfolios designer-developer sites · navigation-examples sidebar and app shell · sources primary specs, the technical floor

agents/ — role prompts for delegated review

design-director visual maturity · ux-architect flows and IA · conversion-designer CTAs and friction · design-critic adversarial banned-pattern scan · frontend-handoff-reviewer spec completeness · accessibility-auditor WCAG pass · performance-auditor CWV pass · copy-editor UI copy · motion-designer motion audit · design-systems-auditor token drift · reference-hunter real examples and competitive analysis

templates/ — output scaffolds

frontend-tz developer handoff · design-system-master the locked MASTER · component-spec one component · design-review-report audit output · project-brief · redesign-brief · ux-audit-report

Directory map

DirectoryHoldsEntry points
rules/22 numbered rules, one concern each00-escalation-protocol · 01-visual-hierarchy · 02-layout-and-grid · 03-typography · 04-color · 05-animation · 06-components · 07-accessibility · 08-performance · 09-responsive · 10-forms · 11-data-tables · 12-admin-panels · 13-saas-products · 14-landing-pages · 15-iconography · 16-design-for-seo · 17-motion-react · 18-css-framework-selection · 19-contrast-standards · 20-rendered-verification · 21-one-shot-build
blueprints/10 step-by-step build protocolslanding-page-from-scratch · interactive-landing-page · saas-app-from-scratch · admin-panel-from-scratch · pricing-page-from-scratch · onboarding-flow-from-scratch · portfolio-from-scratch · ecommerce-from-scratch · website-from-scratch · redesign-existing-page
checklists/6 review gatesglobal-design-review · ui-review · landing-conversion-review · admin-panel-review · frontend-handoff-review · wow-effects-checklist
patterns/7 component groupseffects/ · marketing-blocks/ · product-ui/ · admin-ui/ · content-ui/ · navigation/ · states/
recipes/16 targeted improvementsextract-design-from-reference · create-wow-hero · make-page-more-premium · add-dark-mode · add-page-transitions · improve-* (hero, forms, navigation, typography, mobile, pricing, onboarding, loading, empty states)
industries/13 sector rule setstech-saas · finance · health · education · real-estate · travel · services · b2b-products · b2c-products · content-media · entertainment · government · non-profit — or call GlobalDesignSkill:get_sector_context_tool
tokens/Ready token filesdesign-tokens.json (DTCG) · tokens.css · tokens-dark.css
integrations/Tool-specific setupclaude-code · cursor · windsurf · github-copilot · chatgpt · figma · 21st-dev · frameworks · hyperframes
validators/External check setupaxe-core · lighthouse-ci · bundle-analyzer
evals/The skill's own test suiteagent-cases/ 14 behavioural evals · fixtures/ rendered-verification pages · golden/ reference outputs

Upstream repository, for updates only: https://github.com/staurus86/global-design-skill

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

and 12 more files.

Frequently asked questions

What does the Global Design AI skill do?

Design operating system for web UI/UX work. Use when the user asks to design, build, redesign or audit a web interface: landing pages, SaaS products, admin panels, dashboards, components, forms, animations, color systems, typography, developer handoff specs. Trigger phrases: 'design a page', 'build a landing page', 'create a SaaS UI', 'audit my design', 'review the UI', 'create a component', 'frontend spec', 'improve the interface', 'color tokens', 'add animation', 'looks AI-generated', 'сделай сайт', 'редизайн страницы'. Enforces design-token integrity, eight named quality gates, a banned-...

Why use Global Design on TypingMind?

Because you install it once and use it with any model. Global Design is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Global Design in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/staurus86/global-design-skill/tree/master/skills/global-design. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Global Design?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Global Design?

As many as you like. As long as a model supports skills, you can use Global Design with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Global Design AI skill free?

Yes. It is published on GitHub by staurus86 under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇