domains / design / stitch-design-system-guidance / stitch

stitch

Skill card

combined

The two copies target the same Google Stitch workflow but encode different, non-reproducible procedures. Leonxlnx/taste-skill is a semantic, anti-generic design-taste generator that produces a DESIGN.md with atmosphere, banned AI clichés, typography bans, hero rules, and motion philosophy. bergside/awesome-design-skills is a brand-specific, enterprise drag-and-drop design-system authoring skill with fixed tokens (Ubuntu/Oswald, #072C2C, etc.), accessibility-first rules, and a required output structure. Neither is a shortened or wrapped version of the other; crowning one would hide a real alternative. They agree on the high-level goal (produce Stitch-ready design guidance) and on the need for explicit anti-patterns and component states, but they conflict on palette, typography, tone, and output structure.

Repositories

RepositoryPath
Leonxlnx/taste-skillskills/stitch-skill/SKILL.md
bergside/awesome-design-skillsskills/stitch/SKILL.md

Other copies

RoleRepositoryPath
peerLeonxlnx/taste-skillskills/stitch-skill/SKILL.md
peerbergside/awesome-design-skillsskills/stitch/SKILL.md

Combined skill

---
name: stitch
description: Generate Stitch-ready design-system guidance that is either a semantic anti-generic DESIGN.md or a brand-specific enterprise guideline, depending on the target project.
---

# Stitch Design System Skill

## Purpose
Produce implementation-ready design guidance for Google Stitch. Choose the mode that matches the project:
- **Semantic Taste Mode** (Leonxlnx/taste-skill): for premium, anti-generic, creative interfaces.
- **Enterprise Brand Mode** (bergside/awesome-design-skills): for clean, high-contrast, data-driven drag-and-drop workflows.

## Mode Selection
1. If the user provides a brand, fixed tokens, or an enterprise/data-driven context, use Enterprise Brand Mode.
2. If the user asks for premium, creative, anti-slop, or taste-driven design, use Semantic Taste Mode.
3. If unclear, ask one clarifying question about the intended visual register.

## Semantic Taste Mode
### 1. Define the Atmosphere
- Density: 1–10 (default 4)
- Variance: 1–10 (default 8)
- Motion: 1–10 (default 6)
- Adapt to the user's vibe description.

### 2. Map the Color Palette
- Maximum 1 accent color, saturation below 80%.
- BANNED: AI purple/blue neon, purple button glows, neon gradients.
- Use absolute neutral bases (Zinc/Slate) with high-contrast singular accents.
- Never use pure black (#000000); use Off-Black, Zinc-950, or Charcoal.
- One palette for the entire output.

### 3. Establish Typography Rules
- BANNED: Inter for premium/creative contexts; generic serifs (Times New Roman, Georgia, Garamond, Palatino).
- Preferred: Geist, Outfit, Cabinet Grotesk, Satoshi.
- If serif is needed for editorial/creative: Fraunces, Gambarino, Editorial New, Instrument Serif only.
- Serif is BANNED in dashboards or software UIs.
- Dashboard constraint: Sans-Serif pairings only (Geist + Geist Mono or Satoshi + JetBrains Mono).
- High-density override (density > 7): all numbers use Monospace.
- Body: relaxed leading, max 65 characters per line.

### 4. Define the Hero Section
- Inline Image Typography: embed small contextual photos between words/letters at type-height, rounded.
- No overlapping text or images.
- BANNED: "Scroll to explore", "Swipe down", scroll arrows, bouncing chevrons.
- Centered Hero BANNED when variance exceeds 4; use Split Screen, Left-Aligned, or Asymmetric Whitespace.
- Maximum one primary CTA; no secondary "Learn more" links.

### 5. Describe Component Stylings
- Buttons: tactile push feedback on active; no neon outer glows; no custom cursors.
- Cards: only when elevation communicates hierarchy; tint shadows to background hue; for high-density, use border-top dividers or negative space.
- Inputs: label above, helper text optional, error below.
- Loading: skeletal loaders matching layout dimensions; no generic circular spinners.
- Empty states: composed compositions indicating how to populate data.
- Error states: clear, inline reporting.

### 6. Define Layout Principles
- No overlapping elements; no absolute-positioned content stacking.
- BANNED: generic "3 equal cards horizontally"; use 2-column Zig-Zag, asymmetric grid, or horizontal scroll.
- CSS Grid over Flexbox math; never use calc() percentage hacks.
- Contain layouts with max-width constraints (e.g., 1400px centered).
- Full-height sections: min-h-[100dvh], never h-screen.

### 7. Define Responsive Rules
- Every design must work across all viewports.

### 8. Motion Philosophy
- Specify animation engine, spring physics, and perpetual micro-interactions.

### 9. Anti-Patterns
- Explicit list of banned AI design clichés.

## Enterprise Brand Mode
### Brand
Everything you need – data, apps, and AI in an intuitive drag and drop interface to automate your workflows.

### Style Foundations
- Visual style: clean, high-contrast, enterprise.
- Typography: primary=Ubuntu, display=Oswald, mono=Ubuntu Mono; weights 100–900.
- Color tokens: primary=#072C2C, secondary=#FF5F03, success=#16A34A, warning=#D97706, danger=#DC2626, surface=#EDEADE, text=#111827.
- Spacing: comfortable density mode.

### Accessibility
- WCAG 2.2 AA, keyboard-first interactions, visible focus states.

### Writing Tone
- Confident, helpful, friendly, professional.

### Rules: Do
- Prefer semantic tokens over raw values.
- Preserve visual hierarchy.
- Keep interaction states explicit.

### Rules: Don't
- Avoid low contrast text.
- Avoid inconsistent spacing rhythm.
- Avoid decorative motion without purpose.
- Avoid ambiguous labels.
- Avoid mixing multiple visual metaphors.
- Avoid inaccessible hit areas.

### Guideline Authoring Workflow
1. Restate the design intent in one sentence before proposing rules.
2. Define tokens and foundational constraints before component-level guidance.
3. Specify component anatomy, states, variants, and interaction behavior.
4. Include accessibility acceptance criteria and content-writing expectations.
5. Add anti-patterns and migration notes for existing inconsistent UI.
6. End with a QA checklist that can be executed in code review.

### Required Output Structure
- Context and goals
- Design tokens and foundations
- Component-level rules (anatomy, variants, states, responsive behavior)
- Accessibility requirements and testable acceptance criteria
- Content and tone standards with examples
- Anti-patterns and prohibited implementations
- QA checklist

### Component Rule Expectations
- Define required states: default, hover, focus-visible, active, disabled, loading, error.
- Describe interaction behavior for keyboard, pointer, and touch.
- State spacing, typography, and color-token usage explicitly.
- Include responsive behavior and edge cases (long labels, empty states, overflow).

### Quality Gates
- No rule should depend on ambiguous adjectives alone; anchor each rule to a token, threshold, or example.
- Every accessibility statement must be testable in implementation.
- Prefer system consistency over one-off local optimizations.
- Flag conflicts between aesthetics and accessibility, then prioritize accessibility.

### Example Constraint Language
- Use "must" for non-negotiable rules and "should" for recommendations.
- Pair every do-rule with at least one concrete don't-example.
- If introducing a new pattern, include migration guidance for existing components.

## Conflicts
- **Leonxlnx/taste-skill** bans Inter and generic serifs and mandates unique fonts; **bergside/awesome-design-skills** mandates Ubuntu/Oswald/Ubuntu Mono.
- **Leonxlnx/taste-skill** bans AI purple/blue neon and uses a taste-spectrum atmosphere model; **bergside/awesome-design-skills** uses fixed enterprise tokens and a WCAG 2.2 AA-first structure.
- **Leonxlnx/taste-skill** produces a DESIGN.md with semantic visual descriptions; **bergside/awesome-design-skills** produces a structured guideline with a QA checklist and token-based rules.