Design Craft logo

Design Craft

Community
wasintoh
design-craft

Constraint + process skill for UI that reads "designed", not "AI-generated". Mandatory two-pass process (token plan grounded in the subject's world + ONE signature element, then self-critique) whose output is the project's root DESIGN.md — the design contract re-read on every UI generation. Navigation selection matrix, logo rule (vary expression not placement), icon discipline (one library, labeled), typography pairing recipe, dominant+accent color, copy rules — on top of a non-negotiable usability floor (focus rings, empty/loading/error states, WCAG AA, <=200ms). Ships with AVOID-LIST.md (versioned AI-tell list) and DESIGN-TEMPLATE.md (DESIGN.md template). Triggers: build UI, new component, design review, polish UI, "looks like AI made it", design system, style guide, /toh-vibe, /toh-ui, /toh-design.

Overview

Publisherwasintoh
Repositorytoh-framework
Skill namedesign-craft
Stars
96
Forks
19
Bundled files
2
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.

  • 2 bundled files

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

  • Open source

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

Installation

Install the Design Craft 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.
  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/wasintoh/toh-framework.git /tmp/toh-framework
mkdir -p .claude/skills
cp -r /tmp/toh-framework/src/skills/design-craft .claude/skills/design-craft
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Design Craft 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 Design Craft 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 Design Craft 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.

Design Craft

Two layers — never confuse them:

LayerWhatWhere decided
Creative layerPalette, type, nav, signature element — different per projectRoot DESIGN.md, generated from DESIGN-TEMPLATE.md
Usability floorEvidence-backed rules that never bendBottom of this file

Sibling files (same folder):

  • AVOID-LIST.md — versioned AI-tell list. Consult during identity generation and every design review.
  • DESIGN-TEMPLATE.md — template + filled example for the generated per-project root DESIGN.md.

The contract: root DESIGN.md is the project's design identity. Re-read it before every UI generation — never work from memory of it. If it doesn't exist and UI is about to be written, generate it first (two-pass process below → template). For UI projects it is the first plan task (design-reviewer authors it before any UI work).


Meta-Rule

Defaults are fine when chosen for the brief. Never when inherited.

Inter, indigo-500, rounded-lg, a 3-card feature row — each is legitimate when the brief calls for it, and slop when it appears because it's the training-data median. Every value in the UI must trace to a DESIGN.md token chosen for THIS project.


Two-Pass Process (mandatory before any UI code)

Pass A — compact token plan. Ground it in the subject's actual world — materials, place, era, mood. (A restaurant supply marketplace lives among stainless steel, wet markets, delivery bikes — not "warm and friendly".)

  • 4-6 named hex colors incl. hue-biased neutrals (never pure gray)
  • Display / body / utility typeface roles + weights
  • One-sentence layout concept
  • ONE signature element — the single bold, memorable move (distinctive hero treatment, unusual data display, strong typographic motif). One. Everything else stays quiet.

Pass B — self-critique. Ask: "Would I produce this exact plan for any similar brief?" If yes → revise and state what changed. Check the plan against AVOID-LIST.md.

Only after Pass B survives: write the plan into DESIGN.md and code — deriving every color/type/radius/motion value from it. No raw Tailwind palette colors (indigo-500) that bypass the tokens.


Navigation Selection Matrix

Project typeNavigation pattern
Marketing / landingVisible top bar, 4-7 text links
Dashboard / appLeft sidebar with labeled icons
Content / docsTop bar + secondary sidebar / table of contents
E-commerceTop bar + mega menu for categories
Peer views (same-level sections)Tabs

Hard rules, all patterns:

  1. Never a hamburger on desktop — hiding nav halves discoverability. Visible links at md+ always.
  2. Nav items are real links (<a> / <Link>), never divs with onClick.
  3. URL reflects state — views are shareable and back-button-safe.

Logo Rule — vary expression, not placement

  • Top-left, clickable to home — always, for apps and sites (NNGroup: returning home is ~6x harder with a centered logo).
  • Brand variety comes from wordmark treatment (typeface, weight, spacing), a motif, or an accent system — not from moving the logo.
  • Centered logo ONLY for editorial/luxury briefs that explicitly call for it, and only with a compensating home affordance (visible "Home" link).

Icon Discipline

  • Exactly ONE library per project, declared in DESIGN.md: Lucide / Phosphor / Heroicons / Tabler — plus stroke weight and size grid (e.g. 16/20/24).
  • Every icon gets a text label — except true universals (search, close ×, chevron).
  • Icon-only buttons require aria-label.
  • Never emoji as UI icons.
  • "No icons" is a valid choice — often the stronger one for editorial identities.
  • Icon-tile-above-heading cards are the #1 AI tell (see AVOID-LIST). Icons support content; they don't headline it.

Typography Pairing Recipe

Formula: characterful display + complementary body + optional mono/data face.

  • Inter / Geist are body/utility faces ONLY — never the personality carrier.
  • Use weight extremes: pair 100-200 against 800-900 instead of everything at 400-600.
  • Hero: 3x+ size jump from body. Scale ratio >= 1.25. Body measure 65-75ch. tabular-nums on any column of numbers.
  • Category starting points — rotate; never converge across projects:
    • Editorial / food / lifestyle → Fraunces, Newsreader
    • Technical / data → IBM Plex family
    • Startup / product → Cabinet Grotesk, Satoshi These are prompts, not defaults. Pick for the brief; if the last project used it, pick differently.
  • Thai: IBM Plex Sans Thai / Noto Sans Thai / Anuphan / Sarabun. Next.js: load via next/font (self-host, no layout shift) — never a raw Google Fonts <link>.

Type scale (Tailwind)

Display / Hero:   text-4xl / text-5xl+  display face   tracking-tight
Page Title:       text-2xl  font-semibold              (24px)
Section Title:    text-lg   font-medium                (18px)
Card Title:       text-base font-medium                (16px)
Body:             text-sm   font-normal                (14px, line-height 1.5-1.6)
Small / Caption:  text-xs   text-muted-foreground      (12px)

Rules: headings never font-normal · body regular, never bold-everything · max 3 sizes per component · no long ALL-CAPS runs.


Color

  • Dominant + sharp accent — commit to a dominant character, cut it with one sharp accent. Timid even distribution reads generic.
  • Semantic status colors (good / warn / critical) are separate tokens from the brand accent — never reuse the accent for status.
  • Design light AND dark token-first; both stated in DESIGN.md. Dark mode is its own palette, not inverted gray.
  • Palette sourced from the subject's world or a named cultural aesthetic — name the source in DESIGN.md.
  • Neutrals are hue-biased toward the palette, never pure #808080. No pure #000 text. Contrast passes WCAG AA in both modes.
  • Wire tokens as CSS custom properties (Tailwind 4 @theme / shadcn variables): --background --foreground --muted-foreground --border --primary --destructive map to DESIGN.md tokens.
  • No gradients on primary buttons — solid reads intentional.

Copy Rules

  • Active voice. No buzzwords (streamline, empower, supercharge, unleash).
  • Buttons say exactly what happens: "Save invoice", not "Continue" / "Submit".
  • Errors state the fix: "Card declined — try another card", never an apology or a stack dump.
  • Declare a case system in DESIGN.md (e.g. sentence case everywhere) and keep it.
  • Real content stance: realistic domain data, no Lorem ipsum, no "Welcome back, User! 👋". Copy follows the project language (Thai = real Thai copy, not stiff translation).

Density & Spacing

Layout reflects the real work, not an empty template:

  • Dashboard / admin / data → dense: tables, stat rows, real data. Not "3 floating cards + giant hero".
  • Marketing / landing → narrative sections with substance; hero allowed but earns its space.
  • Form / settings → vertical flow, related fields grouped.
  • Mix composition: table / list / stat / section — never Card → Card → Card. Don't center everything.

Spacing scale (Tailwind — use consistently)

2  (8px)   inline / icon gap
3  (12px)  gap within a component
4  (16px)  standard gap between elements
6  (24px)  between components / card padding
8  (32px)  between sections
12 (48px)  major page sections

Same level = same gap. No random p-3 here, p-5 there.

Border-radius — vary by element, cap declared in DESIGN.md

Inputs / small buttons:  rounded-md   (6px)
Cards / dialogs:         rounded-lg / rounded-xl  (8-12px)
Avatars / pills / badge: rounded-full
Cap:                     12px (rounded-xl) for general containers

Component Patterns (structural — visual character comes from DESIGN.md tokens)

React 19: ref is a normal prop — no forwardRef.

tsx
// Card — thin border + light shadow; no gradient border, no glassmorphism, no rounded-3xl
<Card className="bg-card border shadow-sm">
  <CardContent className="p-6">{/* ... */}</CardContent>
</Card>

// Buttons — one primary per view; no gradient / 3D
<Button>Save invoice</Button>
<Button variant="outline">Cancel</Button>
<Button variant="destructive">Delete</Button>
<Button variant="ghost" size="icon" aria-label="Settings"><Settings className="size-4" /></Button>

// Form (React 19 Actions) — label above input, inline errors
<form action={formAction} className="space-y-4">
  <div className="space-y-2">
    <Label htmlFor="name">Name</Label>
    <Input id="name" name="name" />
    {state?.errors?.name && <p className="text-sm text-destructive">{state.errors.name}</p>}
  </div>
  <SubmitButton /> {/* useFormStatus() → disabled + pending label */}
</form>

// Table — muted header, hover rows, right-aligned actions; no zebra stripes, no centered content
<TableRow className="hover:bg-muted/50">
  <TableCell>John Smith</TableCell>
  <TableCell><Badge>Active</Badge></TableCell>
  <TableCell className="text-right"><Button variant="ghost" size="sm">Edit</Button></TableCell>
</TableRow>

Motion

Motion gives feedback; it is not decoration. Fast (150-200ms), subtle, ease-out.

tsx
<div className="transition-colors hover:bg-muted" />
<Card className="transition-shadow hover:shadow-md" />
// Framer Motion enter: { opacity: 0, y: 8 } → { opacity: 1, y: 0 }, duration 0.2, easeOut
// List stagger: 0.05 max · press: whileTap={{ scale: 0.98 }}

Hover ~150ms · general ~200ms · cap 200ms · no bounce/elastic, no dramatic entrances, no animating everything. Respect prefers-reduced-motion.


NON-NEGOTIABLE USABILITY FLOOR

The creative layer never overrides these:

RuleDetail
Logo top-left, clickable to homeAlways for apps/sites (NNGroup ~6x evidence)
No hamburger on desktopVisible nav links at md+
Icons labeled, one libraryaria-label on icon-only buttons; never emoji as icons
Focus statesfocus-visible:ring-2 ring-ring on every interactive element
Interactions <= 200msease-out; no bounce
WCAG AA contrastLight and dark
Empty statesReal sentence + next action — never bare "No data"
Loading statesSkeletons matching the real layout — not a centered spinner
Error statesInline: what happened + what to do next
Hover / active feedbackEvery button/row (hover:bg-muted, active:scale-[0.98])
Disabled / pendingSubmit buttons show working state (useActionState / useFormStatus)

Review Checklist

  • Root DESIGN.md exists and was re-read for this generation
  • Every color/typeface/radius/motion value traces to a DESIGN.md token
  • Signature element present on the flagship view; everything else quiet
  • Zero AVOID-LIST patterns (scan AVOID-LIST.md groups)
  • Nav matches the matrix · logo top-left + clickable · URL reflects state
  • One icon library, labeled · type pairing not Inter-as-display
  • Usability floor: focus/empty/loading/error/hover/pending all present, WCAG AA
  • Final: would a user tell AI made it? → must be no. Would two different briefs get this same design? → must be no.

Design Craft — one identity per project, chosen on purpose, on top of a floor that never bends.

Bundled files

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

Frequently asked questions

What does the Design Craft AI skill do?

Constraint + process skill for UI that reads "designed", not "AI-generated". Mandatory two-pass process (token plan grounded in the subject's world + ONE signature element, then self-critique) whose output is the project's root DESIGN.md — the design contract re-read on every UI generation. Navigation selection matrix, logo rule (vary expression not placement), icon discipline (one library, labeled), typography pairing recipe, dominant+accent color, copy rules — on top of a non-negotiable usability floor (focus rings, empty/loading/error states, WCAG AA, <=200ms). Ships with AVOID-LIST.md (...

Why use Design Craft on TypingMind?

Because you install it once and use it with any model. Design Craft 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 Design Craft in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/wasintoh/toh-framework/tree/main/src/skills/design-craft. 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 Design Craft?

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 Design Craft?

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

Is the Design Craft AI skill free?

Yes. It is published on GitHub by wasintoh 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 👇