Skill: Advanced Web Research via Jina Reader (r.jina.ai)

Skill: Advanced Web Research via Jina Reader (r.jina.ai)

1. Core Function

Use Jina Reader to convert complex URLs, PDFs, and JS-heavy sites (like Notion, LinkedIn, Twitter) into clean, LLM-friendly Markdown.

2. Usage Patterns

• Basic: Prepend https://r.jina.ai/ to any URL.
• Search: Use https://s.jina.ai/?q=YOUR_QUERY to get a summarized web search result.
• PDF: Prepend the prefix to a direct PDF link for high-quality text extraction.

3. Optimization Headers (for tool calls)

When making a web request, consider these headers to save tokens and improve quality:

• X-Target-Selector: Specify a CSS selector (e.g., article, .main-content) to retrieve ONLY the relevant part.
• X-With-Generated-Alt: Set to true to get AI-generated descriptions for images.
• X-With-Links-Summary: Set to true to get a list of all links at the end of the document (useful for multi-step crawling).
• X-Return-Format: Default is markdown. Use text for even more token savings if formatting isn't needed.

4. When to use Jina instead of standard tools:

1. The site is a Single Page Application (SPA) or requires JavaScript.
2. The site is behind a soft paywall or has aggressive anti-bot measures.
3. The page is extremely long/noisy (Jina's cleaning is superior).
4. You need to "read" a PDF from a URL.
5. You need to see descriptions of images/charts on the page.

5. Implementation Note

If a standard read_web_page call fails or returns "JS required" / "Captcha", immediately retry using the https://r.jina.ai/ prefix.