What problem does it solve? Webpages from platforms like WeChat (mp.weixin.qq.com) and Zhihu often block automated fetching with anti-bot pages, captchas, or dynamic rendering, leaving agents with unusable content. This Skill provides a unified URL-to-Markdown pipeline that probes cheaply first and falls back to the MinerU API, always returning a traceable result with source links. ## Core Features & Use Cases - Two-tier extraction workflow: Probes with low-cost web_fetch first, then falls back to the official MinerU API (MinerU-HTML model) when the probe fails or the domain is whitelisted. - Domain whitelist and heuristics: Ships reference rules for known anti-bot domains (WeChat, Zhihu, Xiaohongshu) and failure signals like captcha text, thin content, or 403 errors. - Traceable result contract: Every run returns a JSON contract with source_url, engine, markdown, artifact paths, and a sources list so downstream skills can verify provenance. - Use Case: Given a WeChat article URL that web_fetch cannot read, the Skill routes directly to MinerU and returns clean Markdown plus the local markdown_path and MinerU zip link for auditing. ## Quick Start Ask the agent to extract this WeChat article URL into Markdown and return the result with its source links.
