Research to Website Builder

You are building a website from research output. This skill guides domain classification, design decisions, and implementation.

Workflow

1. Analyze Research → Identify domains, key insights, data points
2. Load Domain Guidelines → Read relevant files from domains/
3. Load Component Specs → Read from components/ (animations, typography, etc.)
4. Create design-spec.md → Merge guidelines into unified spec
5. Build Website → Implement based on design-spec.md
6. Deploy → Publish and provide live URL

Domain Classification

Classify the research into 1-3 domains based on content nature and presentation needs, not keywords.

Available Domains

How to Decide

Ask these questions about the content:

1. What is the core purpose?

   • Inform investors/professionals → business-financial
   • Present scientific findings → academic-technical
   • Tell a story, evoke emotion, explore culture → art-creative-storytelling
   • Show data trends and metrics → data-analytics
   • Communicate policy to citizens → government-policy

2. How should the audience experience it?

   • Read carefully, reference later → academic-technical
   • Scroll through a narrative journey → art-creative-storytelling
   • Explore data interactively → data-analytics
   • Scan for key decisions → business-financial

3. What visual treatment does it need?

   • Minimal, focused on text → academic-technical
   • Image-rich, atmospheric, animated → art-creative-storytelling
   • Chart-heavy, metrics prominent → data-analytics
   • Conservative, branded → business-financial

Classification Output

Example:

domains:
  primary: art-creative-storytelling   # Drives overall experience
  secondary: data-analytics     # Adds specific capabilities
  tertiary: null                # Optional
rationale: |
  Content explores Earth's landscapes and cultural myths - fundamentally
  storytelling about human connection to nature. Needs immersive visuals
  and narrative flow, not academic structure.

Domain Composition Rules

When multiple domains apply, resolve conflicts:

Required Reference Files

Before creating design-spec.md, READ these files based on classified domains:

domains/{primary-domain}.md      # REQUIRED
domains/{secondary-domain}.md    # If secondary exists
components/animations.md         # If domain uses animations
components/typography.md         # REQUIRED for all
components/color-systems.md      # REQUIRED for all
components/visual-patterns.md    # REQUIRED for image-heavy domains

Design Spec Template

Generate design-spec.md with this structure:

# Design Specification

## Domain Classification
- Primary: {domain}
- Secondary: {domain or none}
- Rationale: {why these domains}

## Design Direction
- Style: {e.g., "Data-driven storytelling with investor-grade polish"}
- Mood: {e.g., "Confident, transparent, forward-looking"}

## Color System
- Background: {hex}
- Primary: {hex}
- Secondary: {hex}
- Accent: {hex}
- Data Palette: [{hex array for charts}]

## Typography
- Display: {font} - {usage}
- Heading: {font} - {usage}
- Body: {font} - {usage}
- Mono: {font} - {usage for data}

## Layout Structure
{Section-by-section breakdown}

## Animation Plan
{Scroll behaviors, transitions, micro-interactions}

## Visual Patterns
{Selected patterns from visual-patterns.md based on domain availability}
- Pattern 1: {name} - {where/how it will be used}
- Pattern 2: {name} - {where/how it will be used}

## Data Visualization Plan
{Charts, metrics, interactive elements}

## Image Requirements
{Hero images, diagrams, icons, plus specific images needed for visual patterns}

Content Integrity Rules

The website must be a complete, accurate presentation of the research:

1. Full Coverage: Extract and present ALL key findings, conclusions, and data points. The website is not a teaser.
2. Data Accuracy: Numbers, statistics, and quotes must be preserved exactly. Never paraphrase data.
3. Structure Mapping: Every major heading in the report should have a corresponding section on the site.
4. Insight Hierarchy: Identify 3-5 most critical insights for prominent treatment (hero sections, pull quotes). Supporting details use scannable formats (cards, lists).

Image Strategy (ALL DOMAINS)

High-quality images are essential for premium websites in ALL domains, not just creative. Use image proactively.

Image Requirements by Purpose

How to Source Images

Always prioritize finding real images through search. Use AI image generation only when a suitable image cannot be found via high-quality image search.

1. Image Search Tool (Primary Choice)

Use image search first for accuracy, realism, and credibility.

• Use image search first for accuracy, realism, and credibility.
• Source high-quality photography for real people, places, products, and events
• Find diagrams, charts, or infographics that explain concrete concepts
• Use official brand, company, or product images when applicable
• Prefer authentic, well-lit, high-resolution visuals
• Default rule: If a relevant, high-quality image exists, do not generate one.

2. Image Generation Tool (Fallback Only)

Use AI image generation only when image search does not yield a suitable result. Appropriate use cases include:

• Abstract or conceptual ideas that lack real-world imagery
• Custom atmospheric backgrounds tailored to a specific theme or mood
• Unique hero visuals when no existing image matches the required tone or composition
• Generation is a last resort, not a replacement for image search.

3. Download Remote Images (IMPORTANT)

Always download remote images to local storage before using them in the website:

• Remote URLs may become unavailable, break, or change over time
• Downloaded images ensure the website remains functional after deployment
• Store images in a local /public/images/ or /assets/ directory
• Use relative paths in the code (e.g., /images/hero.jpg) instead of remote URLs
• This applies to both searched images and any external image sources

Image Placement Guidelines

• Every major section should have at least one visual element (image, chart, or diagram)
• Text-heavy sections need visual breaks every 2-3 paragraphs
• Hero sections require strong, full-bleed imagery
• Transitions between sections can use images to shift mood

Asset Rules

1. Icons: Use professional React icon libraries (Lucide, Heroicons). Never use emojis. Avoid icon abuse—use icons sparingly for clarity, not decoration. Not every label needs an icon.
2. Charts: Use professional libraries (Recharts, Chart.js). Never use raw chart images.
3. Images: Actively source high-quality images. A website without images looks incomplete.
4. Fallbacks: Only if appropriate image truly cannot be found, use abstract shapes, gradients, or typographic treatments.

Brand Color Rules

If the research is about or related to a specific organization/company (e.g., company analysis, stock report, product review, case study):

1. Identify the subject: Is a company/organization the main focus or key subject of the report?
2. Research brand colors: Search for their official brand guidelines, logo colors, or visual identity.
3. Apply brand palette: Use their primary brand color as your primary color. Derive secondary/accent colors that complement it.
4. Maintain consistency: The website should feel like it could belong to or represent that organization.

Examples:

• Stock analysis of Tesla → Use Tesla's red/black brand colors
• Case study on Spotify → Use Spotify's green brand color
• Government policy about NHS → Use NHS blue

Anti-Patterns (AVOID)

• Generic AI aesthetics (purple gradients, dark mode defaults)
• Overused fonts (Inter, Roboto, system fonts)
• Cookie-cutter layouts (hero + cards + footer)
• Decorative-only images
• Animations without purpose
• Charts without context or insight callouts
• Emojis anywhere in the UI
• Icon abuse (icons on every label, decorative icons)
• Paraphrasing or approximating data/statistics
• Ignoring brand colors when research is about a specific company/organization

Ugly Card/Box Patterns (NEVER USE)

• Colored rounded icon containers: Icons inside colored rounded squares (e.g., green box with chart icon)
• Icon on every list item: ✓ checkmarks or ✗ icons on every bullet point
• Semantic color overload: Green = good, red = bad applied to everything (cards, text, icons, backgrounds)
• Pastel tinted card backgrounds: Light green, light red, light blue card backgrounds
• Generic comparison cards: Two side-by-side cards with "pros vs cons" or "good vs bad" layout
• Colored text labels: Bright colored uppercase labels like "SUCCESS", "WARNING", "BUY", "SELL"

Instead:

• Use subtle borders or shadows for card definition
• Use whitespace and typography for hierarchy, not color
• Reserve semantic colors (green/red) for actual data changes only
• Keep backgrounds neutral (white, off-white, light gray)
• Use icons purposefully, not on every element

Tech Stack

• Framework: Next.js (TypeScript)
• Styling: TailwindCSS + shadcn/ui
• Animation: GSAP, Framer Motion, Lenis (smooth scroll)
• Charts: Recharts, Chart.js, D3.js
• Icons: Lucide, Heroicons