News To Markdown

Problem Solved

When an AI Agent receives URLs from Toutiao, WeChat, Zhihu, 36kr, or similar sites, raw fetching often hits anti-bot blocks, dynamic rendering, sidebar noise, and expiring signed image URLs. News To Markdown turns those articles into clean, editable Markdown, reducing downstream cleaning work.

How It Works

The workflow is fetch, extract, and localize:
- Fetch fallback: starts with static fetching and can fall back to Playwright for some dynamic pages.
- Dual extraction: chooses between Mozilla Readability and news-extractor-node, with platform-specific selectors and adaptations for 17 sources.
- Image handling: can download remote images locally, avoiding broken links after signed URLs expire on WeChat, Toutiao, and similar sites.
- Publishing fallback: writes a generated placeholder cover to the cover frontmatter when no cover image is detected, reducing downstream publishing failures.

Boundaries And Notes

It fits article-URL conversion, batch extraction from search results, and noise removal with custom selectors. It is not needed for structured feeds such as GitHub, Reddit, X, or Weibo. Before republishing, preserve source and author attribution and check the site's redistribution policy. For the Skill path, audit the third-party npm package source, and dynamic pages may still require a browser environment.