Website Cloning

Website Cloning Pipeline v2.0

Pipeline completa per analisi e riproduzione pixel-perfect di siti web esistenti.

Architettura Pipeline

URL Input
  │
  ├─── Step 1a: Firecrawl MCP ────────── Markdown + HTML + Brand Identity
  ├─── Step 1b: Design Copier MCP ────── CSS→Tailwind diretto + React skeletons
  │
  ├─── Step 2: Playwright Script v2.0 ── 6 output files (tokens, sections, interactive, structure, screenshots, assets)
  │
  ├─── Step 3: AI Analysis ───────────── Report per-sezione + cross-reference sources
  │
  ├─── Step 4: Component Build ────────── Next.js + Tailwind con per-section accuracy
  │
  └─── Step 5: Screenshot Comparison ──── Pixel diff automatico + QA

Step 1a: Firecrawl — Content + Branding

Firecrawl esegue JS-rendered scraping. SEMPRE usare formats multipli:

MCP Tools Disponibili

Prompt Pattern

1. firecrawl_map(url) → lista completa URL
2. firecrawl_scrape(homepage, formats=["markdown", "html", "branding"])
3. Per pagine chiave: firecrawl_scrape(page, formats=["markdown", "html"])

Output Atteso (con branding)

{
  "url": "https://example.com",
  "markdown": "# Page Title\n\n...",
  "html": "<header>...</header><main>...",
  "branding": {
    "colors": { "primary": "#3b82f6", "secondary": "#10b981", "background": "#fff", ... },
    "fonts": { "heading": "Inter", "body": "Inter", ... },
    "typography": { ... },
    "spacing": { ... },
    "components": { ... }
  },
  "metadata": { "title": "...", "description": "...", "ogImage": "..." },
  "links": ["...", "..."]
}

Step 1b: Design Copier — CSS→Tailwind Diretto

Il Design Copier MCP estrae stili e li converte direttamente a Tailwind o React.

MCP Tools

Workflow Design Copier

1. designcopier_snapshot(url) → {html, styles} della pagina
2. designcopier_extract(html, styles, "tailwind") → mapping Tailwind globale
3. Per sezioni chiave (header, hero, features, footer):
   a. designcopier_snapshot(url, selector="header|section|footer")
   b. designcopier_extract(sectionHtml, sectionStyles, "tailwind")
4. designcopier_apply(styles, "react", "HeroSection") → React skeleton

Output: Mapping CSS→Tailwind per-sezione + componenti React scheletro.

Step 2: Playwright Extraction Script v2.0

Script Node.js migliorato che estrae 6 output files invece di 2.

Esecuzione

# Estrazione base (tokens + sections + interactive + structure + screenshots)
node .Claude/templates/clone-pipeline/extract-design.mjs https://target.com

# Estrazione completa (+ download assets: images, svgs, fonts)
node .Claude/templates/clone-pipeline/extract-design.mjs https://target.com --full

# Output personalizzato
node .Claude/templates/clone-pipeline/extract-design.mjs https://target.com --output ./analisi-sito --full

Output Directory v2.0

clone-output/
├── design-tokens.json        ← Token globali + CSS vars per OGNI selector
├── section-styles.json       ← Token PER-SEZIONE (colori, layout, spacing)
├── interactive-states.json   ← Hover/focus/active states + CSS rules
├── structure.json            ← Albero DOM depth 6 con layout completo
├── screenshots/
│   ├── mobile-375.png
│   ├── tablet-768.png
│   ├── desktop-1024.png
│   └── wide-1440.png
└── assets/                   ← Solo con --full
    ├── images/               ← Max 80 immagini
    ├── svgs/                 ← Max 50 SVG inline
    └── fonts/                ← woff2/ttf intercettati

design-tokens.json Schema v2.0

{
  "colors": {
    "backgrounds": ["#ffffff", "#f8f9fa"],
    "texts": ["#1a1a2e", "#666666"],
    "borders": ["#e5e7eb"],
    "all": ["#ffffff", "#f8f9fa", "#1a1a2e", "..."]
  },
  "typography": {
    "families": ["Inter, sans-serif", "JetBrains Mono, monospace"],
    "sizes": ["14px", "16px", "18px", "24px", "32px", "48px"],
    "weights": ["400", "500", "600", "700"],
    "lineHeights": ["1.5", "1.6", "1.75"],
    "letterSpacings": ["-0.02em", "0.05em"]
  },
  "spacing": {
    "paddings": ["8px", "12px", "16px", "24px", "32px", "48px", "64px"],
    "margins": ["8px", "16px", "24px", "32px"],
    "gaps": ["8px", "12px", "16px", "24px"]
  },
  "borders": {
    "radii": ["4px", "8px", "12px", "9999px"],
    "widths": ["1px", "2px"]
  },
  "shadows": ["0 1px 3px rgba(0,0,0,0.1)", "0 4px 6px rgba(0,0,0,0.1)"],
  "gradients": ["linear-gradient(135deg, #667eea 0%, #764ba2 100%)"],
  "transitions": ["all 0.3s ease", "transform 0.2s ease-out"],
  "animations": [
    { "name": "fadeIn", "duration": "0.5s", "timingFunction": "ease-out" }
  ],
  "cssVariables": {
    "root": {
      "--primary": "#3b82f6",
      "--radius": "0.5rem"
    },
    "bySelector": {
      ":root": { "--primary": "#3b82f6", "--background": "#ffffff" },
      ".dark": { "--primary": "#60a5fa", "--background": "#0f172a" },
      "[data-theme=\"dark\"]": { "--primary": "#60a5fa" },
      "@media(prefers-color-scheme: dark) :root": { "--background": "#0f172a" }
    }
  },
  "breakpoints": [640, 768, 1024, 1280],
  "keyframes": {
    "fadeIn": [
      { "keyText": "0%", "style": "opacity: 0" },
      { "keyText": "100%", "style": "opacity: 1" }
    ]
  },
  "effects": {
    "backdropFilters": ["blur(10px)", "blur(20px) saturate(180%)"],
    "filters": ["blur(4px)", "brightness(0.8)", "drop-shadow(0 4px 6px rgba(0,0,0,0.3))"],
    "transforms": ["rotate(45deg)", "scale(1.1)", "translateY(-50%)"],
    "clipPaths": ["polygon(0 0, 100% 0, 100% 85%, 0 100%)", "circle(50%)"],
    "maskImages": ["linear-gradient(to bottom, black 80%, transparent)"],
    "mixBlendModes": ["multiply", "screen", "overlay"],
    "perspectives": ["1000px"],
    "gradientTexts": [{ "selector": ".hero-title", "gradient": "linear-gradient(90deg, #667eea, #764ba2)" }],
    "textStrokes": [{ "selector": ".outline-heading", "width": "2px", "color": "#000" }],
    "objectFits": [{ "value": "cover", "count": 12 }, { "value": "contain", "count": 3 }],
    "aspectRatios": [{ "value": "16/9", "count": 4 }, { "value": "1/1", "count": 2 }],
    "scrollSnaps": [{ "type": "y mandatory", "align": "start" }],
    "customCursors": ["pointer", "url(custom.cur), auto"],
    "pseudoElements": [
      { "selector": ".decorated::before", "content": "''", "position": "absolute", "background": "linear-gradient(...)" },
      { "selector": ".quote::after", "content": "'\"'", "fontSize": "3rem" }
    ],
    "selectionStyle": { "background": "#3b82f6", "color": "#ffffff" },
    "videoBackgrounds": [
      { "src": "hero-bg.mp4", "poster": "hero-poster.jpg", "autoplay": true, "loop": true, "muted": true }
    ]
  }
}

section-styles.json Schema (NUOVO)

{
  "header.site-header": {
    "index": 0,
    "tag": "header",
    "heading": null,
    "rect": { "top": 0, "height": 64 },
    "tokens": {
      "colors": ["#1a1a2e"],
      "backgrounds": ["#ffffff"],
      "fontFamilies": ["Inter, sans-serif"],
      "fontSizes": ["14px", "16px"],
      "fontWeights": ["500", "600"],
      "paddings": ["12px", "16px"],
      "margins": [],
      "gaps": ["24px"],
      "borderRadii": ["8px"],
      "shadows": ["0 1px 3px rgba(0,0,0,0.1)"],
      "gradients": [],
      "layout": {
        "display": "flex",
        "position": "sticky",
        "flexDirection": "row",
        "justifyContent": "space-between",
        "alignItems": "center",
        "width": 1440,
        "height": 64,
        "maxWidth": "1280px"
      },
      "elementCount": 28
    }
  },
  "section.hero": {
    "index": 1,
    "tag": "section",
    "heading": "Build better products",
    "rect": { "top": 64, "height": 720 },
    "tokens": {
      "colors": ["#ffffff", "#1a1a2e"],
      "backgrounds": ["linear-gradient(135deg, #667eea, #764ba2)"],
      "fontFamilies": ["Inter, sans-serif"],
      "fontSizes": ["18px", "48px"],
      "fontWeights": ["400", "700"],
      "paddings": ["64px", "128px"],
      "margins": [],
      "gaps": ["24px"],
      "borderRadii": [],
      "shadows": [],
      "gradients": ["linear-gradient(135deg, #667eea, #764ba2)"],
      "effects": {
        "backdropFilters": ["blur(10px)"],
        "clipPaths": ["polygon(0 0, 100% 0, 100% 85%, 0 100%)"],
        "hasVideoBackground": true
      },
      "layout": { "display": "flex", "position": "relative", "..." : "..." },
      "elementCount": 42
    }
  }
}

interactive-states.json Schema (NUOVO)

{
  "elements": [
    {
      "tag": "button",
      "class": "btn btn-primary",
      "text": "Get Started",
      "baseState": {
        "color": "#ffffff",
        "backgroundColor": "#3b82f6",
        "borderColor": null,
        "boxShadow": null,
        "transform": null,
        "cursor": "pointer"
      },
      "transition": "all 0.2s ease"
    }
  ],
  "cssRules": {
    "hover": {
      ".btn-primary": { "background-color": "#2563eb", "transform": "scale(1.02)" }
    },
    "focus": {
      ".btn-primary": { "outline": "2px solid #93c5fd", "outline-offset": "2px" }
    },
    "active": {
      ".btn-primary": { "transform": "scale(0.98)" }
    }
  }
}

structure.json v2.0 Improvements

• Depth 6 (era 4) — cattura componenti annidati
• Tutti i tag — non solo semantic, include span, button, picture, figure, svg, video
• Layout completo per flex: flexDirection, justifyContent, alignItems, flexWrap, gap
• Layout completo per grid: gridTemplateColumns, gridTemplateRows, gap, gridAutoFlow
• Position con offsets: position, top, right, bottom, left
• Extra attrs: maxWidth, minHeight, overflow
• Visual effects nel layout: backdropFilter, filter, transform, clipPath, mixBlendMode, objectFit, objectPosition, aspectRatio, opacity, perspective (solo quando presenti)
• Video tag: src, poster, autoplay, loop, muted, playsInline
• Testo catturato per: h1-h6, p (depth ≤3), button, span (depth ≤4, length <60)
• Classi: fino a 8 (era 5)
• Accessibility: role, aria-label
• Picture: sources con srcset

Step 3: AI Analysis

Dopo Step 1 + 2, l'agente @web-analyst produce un report strutturato usando TUTTE le fonti.

Source Priority (quando discordano)

1. Design Copier (designcopier_extract → Tailwind) — più accurato per CSS→class mapping
2. Playwright (section-styles.json, interactive-states.json) — più accurato per computed values
3. Firecrawl branding — utile per validazione incrociata e overview

Report Template v2.0

Vedi 11-web-analyst.md per il formato completo. Il report v2.0 include:

• Per-Section Token Map (non più array piatto)
• Interactive States table
• Dark Mode / Theme Variants
• Cross-Reference Notes

Step 4: Component Build

@component-builder riceve il report e costruisce i componenti Next.js.

Regole di Conversione

Struttura Output Progetto

src/
├── app/[locale]/
│   ├── page.tsx              ← Homepage (clone)
│   └── layout.tsx            ← Root layout con font + theme
├── components/
│   ├── sections/
│   │   ├── hero.tsx
│   │   ├── features.tsx
│   │   ├── pricing.tsx
│   │   └── footer.tsx
│   └── ui/                   ← shadcn components
├── lib/
│   └── animations.ts         ← Framer Motion variants da keyframes estratti
└── globals.css               ← @theme con design tokens mappati (light + dark)

Step 5: Screenshot Comparison

Script automatico per confronto pixel-level.

node .Claude/templates/clone-pipeline/compare-screenshots.mjs \
  --original ./clone-output/screenshots \
  --clone ./Progetti/<slug>/screenshots \
  --output ./clone-output/comparison

Output: comparison-report.json con diff score per viewport e mismatch regions.

Checklist Clone Quality

• Screenshot comparison automatico: diff < 5% per viewport
• Tutti i colori PER-SEZIONE mappati (non solo globali)
• Font corretti (family, weight, size) per heading e body
• Spacing coerente (± 4px tolerance)
• Layout flex/grid corretto (justify, align, wrap, gap)
• Position (sticky, fixed, absolute) con offset corretti
• Hover states implementati per tutti gli elementi interattivi
• Focus states implementati (accessibilità)
• Animazioni riprodotte (timing, easing)
• Responsive breakpoints corretti
• Immagini/SVG posizionati correttamente
• Dark mode implementato (se presente nell'originale)
• Visual effects riprodotti: backdrop-filter, clip-path, blend modes
• Pseudo-elements (::before/::after) implementati dove presenti
• Video backgrounds funzionanti (autoplay, loop, muted)
• Gradient text e text-stroke dove presenti
• Lighthouse score >= originale
• Cross-reference: design-copier e playwright concordano

Troubleshooting