All skills
simota avatar

/vision

@c805268
by shingo imotasimota/agent-skills85 stars
15

Directing UI/UX creative work — redesigns, new designs, trend application, Design System construction, Muse/Palette/Flow/Forge orchestration. Use for design direction. Offers a co-design pair mode.

Use this Skill: https://skilld.dev/gh/simota/agent-skills/vision

This session only. Nothing lands on disk.

referencecomposition-principles.md

≈2k tokens on demand. Your agent reads this file only when SKILL.md points to it.

Composition Principles

Purpose: Use this reference when defining page composition, hero sections, layout structure, or image strategy for any design direction.

Contents

  • First Viewport Rule
  • Hero Contract
  • Layout Restraint
  • Image Strategy
  • Page Structure
  • Composition Audit Checklist

First Viewport Rule

The first viewport is a single composition, not a dashboard.

Principle Rule
Single focus One visual anchor, one message, one action
No dashboard default Never open with cards, stats, or metadata grids
Composition over information density Prioritize visual hierarchy over content volume
Scroll earns complexity Detail and density belong below the fold

Test: Screenshot the first viewport. If it looks like a dashboard or admin panel, it fails.


Hero Contract

Every hero section must satisfy this contract:

Element Required Rules
Brand signal Yes Logo, color, or typography that identifies the brand within 2 seconds
Headline Yes (1 only) Product language, not design commentary. Maximum 12 words
Subline Yes (1 only) Supports the headline with a benefit or clarification. Maximum 25 words
CTA Yes (1 primary) One clear action. Secondary CTA optional but visually subordinate
Visual Yes (1 only) Full-bleed image default. Real product photo preferred over illustration

Prohibited in hero:

  • Card grids or card clusters
  • Statistics or metric displays
  • Metadata or tag lists
  • Multiple competing visuals
  • Decorative gradient backgrounds as the main visual element

Full-bleed image default: Unless the brand direction explicitly requires contained images, hero images should extend to viewport edges.


Layout Restraint

Default: No Cards

Cards are not a layout tool. They are interactive containers.

Use cards when Do NOT use cards when
Content is individually actionable (click, expand, select) Displaying static information sections
Items are peers in a collection (products, team members) Wrapping text blocks for visual separation
Content needs to be compared side by side Creating visual rhythm (use spacing instead)
Content can be reordered or filtered Filling empty space

Section Purpose Rule

One section = one purpose. If a section serves two purposes, split it into two sections.

Anti-pattern Fix
Feature grid + testimonials in one section Separate into Features section + Social Proof section
Pricing table + FAQ combined Separate into Pricing section + FAQ section
Hero with embedded feature highlights Hero section + Features section below

Layout Anti-Patterns

Pattern Problem Alternative
Pill clusters Visual noise, low information density Categorized list or simple text grouping
Stat strips (3-4 numbers in a row) Generic, low-trust without context Inline stats within narrative sections
Icon rows (feature icons in grid) Undifferentiated, forgettable Feature sections with real descriptions
Card grids as default layout Lazy composition, no hierarchy Purpose-driven sections with clear hierarchy
Stacked cards as page structure Everything looks equally important Varied section layouts with visual weight

Image Strategy

Hierarchy of Image Quality

  1. Real product photography — highest trust, strongest brand signal
  2. Custom illustration — brand-aligned, distinctive when well-executed
  3. Curated stock photography — acceptable for supporting sections, not heroes
  4. AI-generated imagery — acceptable for concepts and mood, requires brand review
  5. Generic 3D renders — low trust, high "template" risk
  6. Decorative gradients — never as main visual; acceptable as subtle background treatment

Image Rules

Rule Rationale
Product photos over fake 3D Real imagery builds trust and recognition
No decorative gradients as hero visual Gradients signal "no real content yet"
Images must support the message Every image should reinforce the headline or section purpose
Busy images never behind text Text over complex images fails readability and accessibility
Alt text required for all meaningful images Accessibility baseline

Page Structure

Standard page flow for marketing and product pages:

┌─────────────────────────────┐
│  HERO                       │  Brand + Headline + CTA + Visual
│  (First Viewport)           │  Single composition
├─────────────────────────────┤
│  SUPPORT                    │  Social proof, key benefits,
│  (Trust & Context)          │  or problem statement
├─────────────────────────────┤
│  DETAIL                     │  Features, how-it-works,
│  (Depth & Evidence)         │  use cases, pricing
│  (Multiple sections OK)     │
├─────────────────────────────┤
│  FINAL CTA                  │  Reinforced action with
│  (Closure)                  │  summary value proposition
└─────────────────────────────┘

Structure Rules

Rule Detail
Hero → Support → Detail → Final CTA This order is the default. Deviations require rationale
Each section has one job Name the job. If you can't name it in 3 words, split the section
Visual weight decreases downward Hero is heaviest; Detail sections are lighter
CTA appears twice minimum Once in Hero, once at page end
No orphan sections Every section must connect to the narrative above or below

Intrinsic Layout (2025-2026)

The paradigm has shifted from "Responsive Design" to "Intrinsic Design." A component does not know the viewport size; it only knows its container size.

Container Queries (93.92% global browser support, Dec 2025)

Use @container instead of @media for component-level responsiveness:

.card-wrapper { container-type: inline-size; }

@container (min-width: 400px) {
  .card { display: grid; grid-template-columns: 1fr 2fr; }
}

Fluid Typography with clamp()

Single declaration scales smoothly between min and max:

h1 { font-size: clamp(1.5rem, 4vw, 3rem); }

Intrinsic Sizing with CSS Grid

Content defines layout, not arbitrary breakpoints:

.grid {
  grid-template-columns: repeat(auto-fit, minmax(min(300px, 100%), 1fr));
}

Composition Audit Checklist

Use this checklist when reviewing any page or screen composition:

## Composition Audit

### First Viewport
- [ ] Single composition (not a dashboard)
- [ ] Clear visual anchor
- [ ] One headline, one subline, one CTA
- [ ] Brand identifiable within 2 seconds

### Hero Contract
- [ ] Satisfies all 5 required elements
- [ ] No prohibited elements present
- [ ] Image is full-bleed (or justified deviation)

### Layout
- [ ] No default card grids
- [ ] Each section has one named purpose
- [ ] No pill clusters, stat strips, or icon rows
- [ ] Cards used only for interactive/actionable content

### Images
- [ ] Real product photos preferred
- [ ] No decorative gradients as main visual
- [ ] No busy images behind text
- [ ] All images support the section message

### Page Flow
- [ ] Hero → Support → Detail → Final CTA structure
- [ ] Visual weight decreases downward
- [ ] CTA appears at least twice
- [ ] No orphan sections

Source: SKILL.md on GitHub

No alerts14d5 checks · Risk SAFE
  • Gen Agent Trust Hub14d

    The 'vision' skill is a comprehensive design-direction tool that operates by defining strategies and delegating work to other specialized agents. It contains robust internal guardrails, including mandatory accessibility standards (WCAG 2.2 AA) and ethical AI principles. No security risks were identified.

  • Socket14d

    No alerts

  • Snyk14d

    Risk: LOW · No issues

  • Runlayer6mo

    9 files scanned · No issues

  • ZeroLeaks5mo

    Score: 93/100 · 2 sections analyzed

Signed by skilld at c805268. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub 2 days ago.

Activeupdated 2 weeks ago

README badge

README badge for simota/agent-skills/vision