Frontend Design Review

Frontend Design Review

> Frozen vendored snapshot of microsoft/skills .github/skills/frontend-design-review at e58528d (commit e58528db9a006528a5fb0a2c029790fa6a9a7c0e, MIT) per ADR-0001 declaration→lock→vendored→surfaces. All references are local references/ — no runtime network fetch. See capabilities/upstream.lock design/frontend-design-review and docs/UPSTREAM.md.

Review UI implementations against design quality standards and your design system OR create distinctive, production-grade frontend interfaces from scratch.

Two Modes

Mode 1: Design Review

Evaluate existing UI for design system compliance, three quality pillars (Frictionless, Quality Craft, Trustworthy), accessibility, and code quality.

Mode 2: Creative Frontend Design

Create distinctive interfaces that avoid generic "AI slop" aesthetics, have clear conceptual direction, and execute with precision.

---

Creative Frontend Design

Before coding, commit to an aesthetic direction:

• Purpose: What problem does this solve? Who uses it?
• Tone: minimal, maximalist, retro-futuristic, organic, luxury, playful, editorial, brutalist, art deco, soft/pastel, industrial, etc.
• Constraints: Framework, performance, accessibility requirements.
• Differentiation: What makes this distinctive and context-appropriate?

Aesthetics Guidelines

• Typography: Distinctive fonts that elevate aesthetics. Pair a display font with a refined body font. Avoid Inter, Roboto, Arial, Space Grotesk.
• Color & Theme: Cohesive palette with CSS variables. Dominant colors + sharp accents > timid, evenly-distributed palettes.
• Motion: CSS-only preferred. One well-orchestrated page load with staggered reveals > scattered micro-interactions.
• Spatial Composition: Asymmetry, overlap, diagonal flow, grid-breaking elements, generous negative space OR controlled density.
• Backgrounds: Gradient meshes, noise textures, geometric patterns, layered transparencies, dramatic shadows, grain overlays.

AVOID: Overused fonts, cliched color schemes, predictable layouts, cookie-cutter design without context-specific character.

Match implementation complexity to vision. Maximalist = elaborate code. Minimalist = restraint and precision.

---

Design Review

Design System Workflow

Before implementing:

1. Review component in your Storybook / component library for API and usage
2. Use Figma Dev Mode to get exact specs (spacing, tokens, properties)
3. Implement using design system components + design tokens

During review:

1. Compare implementation to Figma design
2. Verify design tokens are used (not hardcoded values)
3. Check all variants/states are implemented correctly
4. Flag deviations (needs design approval)

If component doesn't exist:

1. Check if existing component can be adapted
2. Reach out to design for new component creation
3. Document exception and rationale in code

Review Process

1. Identify user task
2. Check design system for matching patterns
3. Evaluate aesthetic direction
4. Identify scope (component, feature, or flow)
5. Evaluate each pillar
6. Score and prioritize issues (blocking/major/minor)
7. Provide recommendations with design system examples

Core Principles

• Task completion: Minimum clicks. Every screen answers "What can I do?" and "What happens next?"
• Action hierarchy: 1-2 primary actions per view. Progressive disclosure for secondary.
• Onboarding: Explain features on introduction. Smart defaults over configuration.
• Navigation: Clear entry/exit points. Back/cancel always available. Breadcrumbs for deep flows.

---

Quality Pillars

1. Frictionless Insight to Action

Evaluate: Task completable in ≤3 interactions? Primary action obvious and singular?

Red flags: Excessive clicks, multiple competing primary buttons, buried actions, dead ends.

2. Quality is Craft

Evaluate:

• Design system compliance: matches Figma specs, uses design tokens
• Aesthetic direction: distinctive typography, cohesive colors, intentional motion
• Accessibility: Grade C minimum (WCAG 2.1 A), Grade B ideal (WCAG 2.1 AA)

Red flags: Generic AI aesthetics, hardcoded values, implementation doesn't match Figma, broken reflow, missing focus indicators.

3. Trustworthy Building

Evaluate:

• AI transparency: disclaimer on AI-generated content
• Error transparency: actionable error messages

Red flags: Missing AI disclaimers, opaque errors without guidance.

---

Review Output Format

See [references/review-output-format.md](references/review-output-format.md) for the full review template.

Review Type Modifiers

See [references/review-type-modifiers.md](references/review-type-modifiers.md) for context-specific review focus areas (PR, Creative, Design, Accessibility).

Quick Checklist

See [references/quick-checklist.md](references/quick-checklist.md) for the pre-approval checklist covering design system compliance, aesthetic quality, frictionless, quality craft, and trustworthy pillars.

Pattern Examples

See [references/pattern-examples.md](references/pattern-examples.md) for good/bad examples of creative frontend and design system review work.

---

Acknowledgments

Creative frontend principles inspired by Anthropic's frontend-design skill. Design review principles and quality pillar framework created by @Quirinevwm for systematic UI evaluation.

---

Toolkit integration

Delegation — visual critique vs project health

| Concern | Owner skill | What it does | When to use | |---------|-------------|--------------|-------------| | Project health (tech/mgmt) | technical-unit-assessment + project-assessment-evidence | Evidence-based 1–5 scoring of repo/platform/infra units; no visual judgment | Repo health, CI, security posture | | Design orchestration | design-assessment (planned #373, per ADR-0002 Option A) | Orchestrates discovery→inspect→capture→visual/UX/a11y/responsive/DS/AI-slop/perf→findings→roadmap; reuses evidence semantics; fans out to reviewers | Full product design audit | | Procedural visual critique | frontend-design-review (this skill) | Checklist evaluation of hierarchy, design-system compliance, three pillars (frictionless / quality craft / trustworthy), a11y, responsive, polish, AI-slop — outputs severity-ranked findings with citations | PR review, component review, creative frontend kickoff | | Distinctive visual direction | frontend-design | Studio-lead aesthetic choice, palette/type/layout thesis | New UI creation with intentional style |

This skill does not score project health. For project health use technical-unit-assessment. For full design product audit, design-assessment will delegate to this skill for the procedural visual pass, alongside web-design-guidelines (Web Interface Guidelines rules) and browser/accessibility/Figma reviewers.

Evidence-cited findings — no fake scores

• Never emit numeric scores (e.g. 72/100) without observable evidence. Use severity bands Blocking / Major / Minor with confidence and evidence per finding, reusing project-assessment-evidence semantics where applicable.
• Every issue must cite: observation (what you saw — screenshot region, code line, token mismatch), impact (user/task consequence), effort, and recommended fix with design-system link or token reference.
• Mark indicators as Not assessed when screenshot/code unavailable rather than inventing a rating. Vision harness unavailable → text-only heuristic review with lower confidence, explicitly flagged.

Wiring to design-assessment (planned)

This skill is a delegate of design-assessment. The orchestrator (see #373) will call it during the VISUAL REVIEW phase, passing captured screens/states (Playwright) and design-system manifest (Figma tokens / Storybook / Tailwind). It runs in parallel with accessibility / web-design-guidelines reviewers where the harness supports subagents, otherwise sequentially. Future design-assessment will document the full routing.

Provenance

• Upstream: microsoft/skills .github/skills/frontend-design-review + references/* at e58528db9a006528a5fb0a2c029790fa6a9a7c0e (MIT), LICENSE vendored here as LICENSE.
• Lock: capabilities/upstream.lock design/frontend-design-review provenance_digest (commit + content_checksum + license), trust.reviewed_provenance binding.
• References vendored: references/review-output-format.md, references/review-type-modifiers.md, references/quick-checklist.md, references/pattern-examples.md — byte-identical to upstream at pinned commit.
• Update: pull-request per resolved_at; never fetch main at runtime.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: ulises-jeremias
• Source: ulises-jeremias/agent-toolkit
• License: MIT
• Homepage: https://ulises-jeremias.github.io/agent-toolkit-archive/

Install and usage instructions live in the source repository linked above.