AgentSkillsCN

reader-jina

将任意网页以纯净的 Markdown 格式呈现。当 WebFetch 在高 JS 密集型网站上失效，或当用户要求阅读需要 JavaScript 渲染的页面时，可使用此功能。触发条件包括：(1) WebFetch 返回空内容或内容损坏；(2) 用户分享来自高 JS 网站（如单页应用、仪表盘）的 URL；(3) 用户说“读一读这个页面”“这个页面上有什么”。返回的页面内容将以 Markdown 格式呈现。

SKILL.md
--- frontmatter
name: reader-jina
description: "Read any web page as clean Markdown. Use when WebFetch fails on JS-heavy sites, or when the user asks to read a page that requires JavaScript rendering. Trigger when: (1) WebFetch returns empty/broken content, (2) user shares a URL from a JS-heavy site (SPA, dashboards), (3) user says 'прочитай страницу', 'что на этой странице'. Returns page content as Markdown."

Jina Reader

Read any web page as clean Markdown via Jina Reader API (r.jina.ai). Fallback for when WebFetch fails on JS-heavy sites.

Usage

bash
uv run scripts/fetch_page.py "URL"
uv run scripts/fetch_page.py "URL1" "URL2"

The Iron Law

Try WebFetch first. reader-jina is a fallback, not a replacement. Use only when WebFetch returns empty, broken, or incomplete content.

How It Works

Prepends https://r.jina.ai/ to the target URL. Jina renders the page (including JavaScript) and returns clean Markdown.

Limitations

  • •Free tier: 200 requests per minute (no API key needed)
  • •Public pages only (no authentication passthrough)
  • •Timeout: 30 seconds (some heavy pages may fail)
  • •Content may be summarized for very large pages

Related Skills

  • •reader-twitter — specialized skill for Twitter/X posts and threads
  • •reader-youtube — transcript extraction from YouTube videos