← ClaudeAtlas

web-site-to-documentlisted

Convert a public website into one structured document (Word, PDF, or Markdown) — a whole site, one section, or a single page. Use to archive a URL, or to scrape a JavaScript-rendered site.
StefanoZaghi1987/ClaudeSkills · ★ 2 · Data & Documents · score 73
Install: claude install-skill StefanoZaghi1987/ClaudeSkills
# web-site-to-document This skill scrapes any public website (including all linked subpages) and converts the full content into a single structured document: Word (.docx), PDF (.pdf), or Markdown (.md). ## Execution branches Pick the branch that matches the environment: - **If this environment provides the built-in document skills (claude.ai):** collect the pages with the environment's built-in web fetch. For JavaScript-rendered sites, use the Chrome MCP connector if the user has one (workflow: `references/chrome-mcp-extraction.md`) — otherwise warn the user that extraction may be limited (see Step 2). Build the file with the environment's document skills: for `.docx` and `.pdf`, read and follow their documentation when the platform exposes it; for `.md`, write the file directly. Let the platform's file delivery present the file in the conversation; do not construct file paths or download links yourself. - **If it does not (Claude Code, local Python):** run the bundled pipeline, on Python 3.10 or newer. The needed packages are `requests beautifulsoup4 lxml` (always), plus `python-docx` for Word. For PDF there are two engines: LibreOffice headless is preferred when installed, and it builds on `python-docx`; when LibreOffice is absent, `weasyprint` is needed instead and needs no `python-docx`. On Windows weasyprint needs the GTK runtime, which `pip` does not install — `pip install weasyprint` succeeds and the import still fails — so LibreOf