1:1 Website Clone Skill — Pixel-Perfect Full-Site & Full Produc…

1:1 Website Clone Skill — Pixel-Perfect Full-Site & Full Product Catalog Clone Master Skill

Positioning: Concentrates "forensic-grade DOM/CSS collection + Akamai/Cloudflare real-device penetration + full-category product crawler with local asset storage + 1:1 interactive store reconstruction + automated layout/SEO/multi-viewport visual QA" into a single closed-loop ultimate 1:1 website clone skill.

---

I. Skill Fusion Architecture (6 Underlying Capabilities United)

This skill merges the core capabilities of 6 local and GitHub top skills, solving the pain points that a single skill cannot address: "high-protection website penetration", "lazy-load original image extraction", "full-category product batch import", and "pixel-perfect static/interactive restoration":

Module Source	Core Responsibility	Fixed Phase in This Skill
1. Jane-xiaoer/claude-skill-web-clone	5-phase forensic workflow, DOM semantic tree mapping, block-level 1:1 visual restoration	Phase 1 (DOM block mapping) & Phase 7 (multi-scroll-position screenshot comparison)
2. ericshang98/perfect-web-clone-skill	Browser getComputedStyle precise token extraction, DESIGN.md contract, sticky navigation state machine	Phase 1 (CSS precise extraction) & Phase 4 (design specification synthesis)
3. ego-browser (Real Chromium Engine)	Penetrate Akamai (HTTP 403) / Cloudflare, automatically close lock-screen modals, browser context page.fetch download	Phase 1 ~ Phase 3 (full real-device environment collection and zero-403 asset storage)
4. renwork-web-create-skill (Mode B v3)	Zero Placeholder Policy, automated SEO/GEO (sitemap.xml/robots.txt/llms.txt)	Phase 2 (100% local assets) & Phase 6 (SEO/GEO generation)
5. wideway-export-site-builder-skill	Full product list page (PLP), split-screen product detail page (PDP), shopping bag, wishlist, and front-end product upload CMS	Phase 5 (full-site product interactive store and online upload engine)
6. layout-typography-auditor	Single <h1> validation, semantic interactive tags (no <div onclick>), form autocomplete, :focus-visible audit	Phase 6 (static code quality and accessibility 0-error gate)

---

II. Iron Rules and Red Lines (Non-Negotiable Rules)

1. Strictly forbid direct curl / urllib bare scraping of high-protection sites: When encountering Akamai (e.g., us.louisvuitton.com), Cloudflare, or other protections returning 403 Forbidden, you must go through ego-browser's real-device page for DOM extraction and page.fetch(url, { saveAs }) local saving.
2. Modal & Scroll Unlock First: Upon first opening the target site, you must first detect and click to close region selection modals (e.g., .lv-modal__container button.lv-modal__close), cookie overlays, or subscription popups, and verify that window.scrollTo(0, 500) actually changes window.scrollY, to prevent capturing locked-screen semi-transparent overlays or blank lazy-load areas.
3. Penetrate lazy-load extraction of 4K/HD original images: For <picture> dynamic data-srcset or <noscript><img src="..."></noscript>, you must write scripts to preferentially parse high-res source URLs (e.g., ?wid=2048 / ?wid=1090), strictly forbidding the capture of 1x1 base64 placeholders.
4. 100% Localized Assets and Full Product Import: All homepage posters, category covers, and full-category product studio white/off-white background images (PNG/JPG) must be downloaded locally to assets/images/ and assets/products/; all official icons and logos must be inline original <svg viewBox="...">, strictly forbidding the use of any external hotlink URLs or generic image libraries (Unsplash/Pexels).
5. Isolated Workspace Protection: If the current workspace already has existing business code, the complete project must be built in an independent subdirectory (e.g., <brand>-clone/), strictly forbidding overwriting existing files in the root directory.

---

III. Standard 7-Step Pixel-Perfect Clone Full Process (7-Phase Pipeline)

Phase 1: Real-Device Forensic Reconnaissance and Modal Penetration (scripts/recon_and_harvest.mjs)

1. Start ego-browser TaskSpace, set desktop viewport 1440×900.
2. Navigate to the target URL, automatically detect and close Country/Region modals and Cookie banners, remove body scroll lock.
3. Scroll in segments to trigger full-page lazy loading, extract:
   • RECON/original-recon.json & RECON/homepage-blocks.json: titles, subtitles, CTA links, grid column counts, and exact coordinates for all sections on the page.
   • RECON/css-spec.json: extract precise font size, line height, letter spacing, padding, background gradients, and border colors for Header, Hero, Section titles, Product Cards, Footer via getComputedStyle.
   • RECON/svgs.json: extract all original vector SVGs (brand wordmark, search, wishlist, shopping bag, close, arrow, social icons).
4. At fixed step intervals along the page height (e.g., 0, 720, 1660, 2450, 3150, 3920, 4630, 5400, 6100), use page.cdp("Page.captureScreenshot") to capture high-definition baseline images without modals at RECON/screenshots/original-clean-{y}.png.

Phase 2: Full-Page High-Definition Media Asset Download (assets/images/)

1. Parse all Hero hero images, category grid images, featured single-item transparent PNGs, and brand story vertical images on the homepage.
2. Within ego-browser, invoke page.fetch(url, { saveAs }) to save all images to assets/images/, verifying each file's byte size is > 1KB.

Phase 3: Full-Site Multi-Category Product Crawler and Batch Upload Import (scripts/crawl_all_categories.mjs)

1. Extract all core category URLs associated with the homepage and navigation menu (e.g., womens-bags, womens-slg, womens-accessories, womens-shoes, mens-bags, mens-slg, mens-shoes, mens-accessories, seasonal limited series, etc.).
2. Sequentially visit each category landing page within ego-browser, automatically extracting from each product card:
   • sku (unique product code)
   • name (official product name)
   • price & priceValue (formatted price and numeric value)
   • category & categoryName (category ID and display name)
   • gender (target audience)
   • image (high-definition studio product image URL)
   • description / material / dimensions (craft description and specifications)
3. Deduplicate across categories by sku, invoke page.fetch to download all product high-definition images to assets/products/prod_<SKU>.png.
4. Automatically compile and generate a structured product database:
   • data/products.json (standard JSON database)
   • js/products-data.js (front-end zero-latency loading global variable window.LV_CATALOG_DATA / window.SITE_CATALOG_DATA)

Phase 4: Design Specification Document Generation (DESIGN.md & NOTES.md)

Solidify collected visual parameters into DESIGN.md:

• Color Tokens: main text color (e.g., #1a1a1a / #000000), secondary text color (#727272), light border (#e1e1e1), product card stone studio background gradient (linear-gradient(180deg, #efeeec 0%, #f6f5f3 50%, #fbfaf9 100%)).
• Typography Scale: custom brand sans-serif font stack, H1/H2 (28px/36px or 24px/32px, letter-spacing: 0.4px), pretitle (10px/16px, uppercase, letter-spacing: 1px), product title and price (14px/20px).
• Layout & Grid: Header height (88px, left/right padding 48px), 4-column product/category grid (grid-template-columns: repeat(4, 1fr), gap: 32px 16px, left/right margin 64px), 3-column brand culture grid (repeat(3, 1fr)).

Phase 5: 1:1 Pixel-Level Page Reconstruction and Full-Featured Product Store (index.html, css/, js/)

1. Homepage 1:1 Visual Reconstruction (index.html + css/lv-styles.css) :
   • Build DOM strictly following original site Section order, aspect ratios (16:9 hero images, 4:5 product and category cards), and inline SVGs.
   • Implement sticky navigation state machine: above the hero section on first screen, transparent background with white text/icons; when scrolling past a threshold (60px) or on mouse hover, smoothly switch to white background with black text/icons and a thin bottom border.
2. Full Product Store and Interaction Engine (js/lv-app.js) :
   • Full Product Catalog View (#lv-catalog-view): Support one-click viewing of all uploaded products (e.g., 86+ items), provide category pill filter bars, price high-to-low sorting, and a one-click return to homepage.
   • Split-Screen Product Detail Modal (PDP): Click any product to display a large high-definition image stage on the left, with SKU number, price, craft/material description, and a Place in Cart add-to-cart button on the right.
   • Real-Time Search Drawer (Search Overlay): Support real-time filtering by product name, SKU, or category, instantly displaying thumbnail results.
   • Persistent Wishlist & Shopping Bag: Persist favorite and add-to-cart states based on localStorage, update top navigation badge counts in real time.
   • Online Product Upload Backend (+ Upload Product CMS Drawer): Support directly filling in new product name, SKU, price, category, image path, and description on the web page, immediately publishing to the full-site catalog with persistent storage.

Phase 6: Static Code, Layout Accessibility, and SEO/GEO Automation Audit (scripts/)

1. Run python3 scripts/seo_geo_generator.py <project_dir> <base_url> to automatically generate sitemap.xml, robots.txt, and llms.txt containing full product URLs.
2. Run python3 scripts/layout_typography_auditor.py <project_dir> to execute static quality gate:
   • Ensure the entire page has only one main <h1>;
   • Ensure all click events are bound to <button> or <a> (or via JS event delegation), eliminating non-semantic <div onclick>;
   • Ensure all text input fields have an autocomplete attribute and an accessible <label>;
   • Ensure all interactive controls have a :focus-visible style, achieving 0 blocking errors, 0 interface guidelines.

Phase 7: Multi-Viewport Real-Device Visual QA and Delivery Report (scripts/visual_qa_verify.mjs)

1. Start a local HTTP preview service, use ego-browser to visit the local clone site.
2. Force full image loading, verify homepage images and full catalog product images have loaded === total and brokenImages.length === 0.
3. Capture full-process comparison screenshots on desktop (1440×900) and mobile (390×844) viewports (homepage scroll positions, full product catalog page, PDP detail page, CMS upload drawer, mobile first screen), verify mobile scrollWidth <= innerWidth (zero horizontal overflow).
4. Close ego-browser TaskSpace (await task.finish({ keep: [] })), generate CLONE_REPORT.md and deliver.

---

IV. Supporting Automation Script List (scripts/)

• scripts/recon_and_harvest.mjs: Real-device modal penetration, DOM/ComputedStyle/SVG extraction, original-site segmented baseline screenshots, and homepage high-definition image downloader.
• scripts/crawl_all_categories.mjs: Full-category product recursive crawling, cross-category SKU deduplication, batch local download of high-definition product images, and products.json / products-data.js compiler.
• scripts/seo_geo_generator.py: Based on data/products.json, automatically generates sitemap.xml, robots.txt, and llms.txt.
• scripts/layout_typography_auditor.py: HTML/CSS semantics, heading hierarchy, form and focus accessibility zero-error audit tool.
• scripts/visual_qa_verify.mjs: Local server auto-mount + ego-browser multi-viewport screenshot comparison and broken link/overflow auto-verification script.