Create distinctive, production-grade frontend interfaces with high design quality.
Use when building UI components, pages, or applications.
Guides from context gathering through implementation to quality gate.
Triggers: "frontend design", "build UI", "design this page", "make it look good"
Installation
Mit Codex oder Claude installieren Kopieren Sie diesen Prompt, fügen Sie ihn in Codex, Claude oder einen anderen Assistant ein und lassen Sie die Skill-Seite prüfen und installieren.
Create distinctive, production-grade frontend interfaces with high design quality.
Use when building UI components, pages, or applications.
Guides from context gathering through implementation to quality gate.
Triggers: "frontend design", "build UI", "design this page", "make it look good"
Use when building UI, pages, or components. Guides context gathering → implementation → frontend-designer-agent quality gate at 7+/10. Triggers: "frontend design", "build UI", "design page".
Status protocol — end every session with one of: DONE (evidence provided) · DONE_WITH_CONCERNS (list each) · BLOCKED (state what blocks you) · NEEDS_CONTEXT (state what you need).
Auto-advance — pipeline: THINK → PLAN → REVIEW → BUILD → VERIFY → RELEASE. Only human gate is spec approval at THINK. On DONE at other stages, print [STAGE] DONE -> advancing to [NEXT-STAGE] and invoke the next skill. On any non-DONE status at any stage, STOP.
Output directory — all artifacts go in docs/superomni/<kind>/<kind>-[branch]-[session]-[date].md. See CLAUDE.md for the full directory map.
TACIT-DENSE — before high-tacit decisions, classify D1 (domain expertise) · D2 (user-facing UX) · D3 (team culture) · D4 (novel pattern). On hit, output TACIT-DENSE [D#]: [question] — My default: [recommendation]. See reference for actions.
Anti-sycophancy — take a position on every significant question. Name flaws directly. No filler ("that's interesting", "you might consider", "that could work").
Telemetry (local only) — at session end, log bin/analytics-log. Nothing leaves the machine.
Goal: Build frontend interfaces that are functional, production-ready, AND visually distinctive. Avoid generic AI aesthetics — every design choice must be intentional.
Iron Law: Never Ship UI Without the Quality Gate
No frontend code ships without running the frontend-designer agent quality gate. A passing score on all 10 dimensions (7+/10 each) is required. Two auto-retry attempts are allowed before escalating to the user.
The Core Problem
Every LLM learned from the same generic templates. Without guidance, you get:
Inter font, purple gradients, cards nested in cards
Gray text on colored backgrounds, glassmorphism everywhere
Rounded rectangles with drop shadows (the "SaaS dashboard" look)
Hero metric layouts, neon accents, gradient text as decoration
The quality test: If someone saw this interface and said "AI made this," would they believe it immediately? If yes — that's the problem.
Phase 1: Context Gathering (mandatory)
You cannot infer design context from code alone. Ask the user ONE question at a time:
Required understanding before proceeding:
Target audience — who uses this? (technical vs non-technical, expert vs beginner)
Brand personality — what feeling should this evoke? (serious, playful, minimal, bold)
Use context — how will it be used? (quick glance vs deep focus, desktop vs mobile, high stress vs relaxed)
Existing design system — are there brand colors, fonts, or components to reuse?
If a config file exists, load it and skip questions it already answers.
Phase 2: Design Direction
Commit to a bold aesthetic direction. Present 3 curated options + Other:
Design Direction Options:
A) Minimalist Precision — Swiss-inspired, generous whitespace, one accent color,
strong typography hierarchy. Think: Linear, Stripe, Vercel.
Pro: Timeless, professional. Con: Can feel cold.
B) Warm Organic — Rounded shapes, earth tones or pastels, hand-drawn accents,
friendly micro-interactions. Think: Notion, Loom, Figma.
Pro: Approachable, memorable. Con: Harder to maintain consistency.
C) Bold Maximalist — High contrast, dramatic typography, intentional color blocking,
confident whitespace. Think: Bloomberg, Pentagram, Readymag.
Pro: Distinctive, authoritative. Con: Polarizing.
D) Other — describe your own aesthetic direction: ___________
Key principle: Bold maximalism and refined minimalism both work — the key is intentionality, not intensity.
Phase 3: Reference Loading
Load relevant reference files on-demand based on the current task. Maximum 2-3 at a time to manage context.
Reference Selection Matrix
Task involves...
Load these references
Text layout, fonts, headings
reference/typography.md
Colors, palette, dark mode
reference/color-and-contrast.md
Layout, spacing, grid
reference/spatial-design.md
Animations, transitions
reference/motion-design.md
Forms, buttons, states
reference/interaction-design.md
Mobile, responsive, breakpoints
reference/responsive-design.md
Labels, errors, microcopy
reference/ux-writing.md
# Load references on-demandcat skills/frontend-design/reference/typography.md # only when needed
If the user requests a reference not yet loaded, load it. If context pressure is high, summarize the current reference into 5 key rules and unload the raw content.
External site DESIGN.md files are treated as brand style input, not as implementation truth. Engineering and accessibility rules must still come from the core references.
Load one brand file from reference/design-md-whitelist.md
Load 1-2 core reference files from the matrix above
Hard limits:
Never load more than one brand DESIGN.md in a session
Never replace quality gate dimensions with brand rules
Keep total active references to 2-3 files
Governance and legal boundary:
Borrow style system patterns, not proprietary identity assets
Do not replicate logos, trademarks, unique illustrations, proprietary copy, or exact layout signatures
Target: "same design language", never "1:1 visual clone"
Naming boundary:
This project already has docs/DESIGN.md (architecture). Keep external style references under skills/frontend-design/reference/ and distilled outputs under docs/superomni/style-profiles/.
Phase 3.5: Style Distillation Output (for reuse and audit)
After implementing with one brand reference, distill reusable style guidance:
Use docs/superomni/style-profiles/design-md-distillation-template.md as the schema
If the template file is unavailable, fallback to sections in reference/design-md-adaptation.md and structure output into: Metadata, Transferable Rules, Quality Gate Mapping, Anti-Copy Constraints, Prompt Fragment; prioritize quality-gate/accessibility constraints over brand-style phrasing.
Distillation rule:
Keep only transferable tokens/rules (typography scale, spacing rhythm, color roles, state behavior, motion principles)
Remove brand-locked identity elements and copyrighted copy
Keep the profile concise and implementation-oriented
Phase 4: Implementation
Build with the loaded reference guidance active. Every implementation must:
Match code complexity to aesthetic vision — a minimalist design needs clean, simple code; a maximalist design may need more CSS
Use modern CSS — prefer oklch(), clamp(), container queries, :focus-visible, prefer-reduced-motion
Design all states — default, hover, focus, active, disabled, loading, error, success
Apply the reference rules — don't just read them, apply them to every decision
Test the squint test — blur your eyes: can you still see the hierarchy?
Responsive: mobile-first, container queries where appropriate
Accessibility: WCAG AA contrast, keyboard nav, focus indicators, screen reader support
Copy: specific labels, 3-part error messages, helpful empty states
Phase 5: Quality Gate
Run the frontend-designer agent for a full design review. Quality gate must clear all 10 dimensions at ≥7/10 (3-attempt auto-retry, then escalate).
Reference: see reference/quality-gate.md for full Gate Protocol (3-attempt retry policy + escalation message) and the canonical Quality Gate Output block.
