Doubao Share to Word

Problem

Doubao share links are usually read-only in the browser, and their content is scattered across SSR data, scripts, and the rendered DOM. Copy-paste often loses code blocks, tables, images, LaTeX, and speaker structure. This skill converts share pages into editable Word documents, focusing on preserving full content and structure.

How It Works

• Accepts doubao.com/thread/..., doubao.com/chat/share?shareId=..., doubao.com/bot/..., or cloud-drive share links, with automatic link-type detection.
• Parses embedded page data first, falling back through modern-run-router-data-fn, modern-run-window-fn, window._ROUTER_DATA, __NEXT_DATA__, and direct DOM parsing.
• For newer content JSON arrays, maps block_type values to text, inline code, full code, and images, then decodes nested JSON payloads.
• Detects speaker roles via role, user_type, ext.agent_id, and alternating message order.
• Outputs a Word document with plain text, Markdown, code blocks, tables, images, LaTeX, quotes, and lists, plus role labels and separators.

Boundaries and Caveats

• Most share pages require JS rendering; if static extraction fails, use Playwright mode or provide rendered HTML.
• Failed image downloads leave URL placeholders without blocking export, and malformed messages are skipped.
• Links without embedded data, changed page structure, or login requirements may need manually saved HTML for parsing.