Sitepanda (Web Scraping Tool)

Instructions

1. When the user provides a URL or asks for website content, use Sitepanda to scrape the page.

2. By default, use the following command to scrape a single page:

   sitepanda scrape <URL> --silent --limit 1

3. If you need to perform recursive scraping (following links), you must ask the user for confirmation before starting, as it may take a long time.

4. Capture the output, which is returned in Markdown format.

5. Read and analyze the extracted content.

6. Respond to the user using only the relevant information from the page.

7. If the content is long, summarize or extract only the necessary sections.

Examples

Example 1

User request: "Please summarize the article at https://example.com/blog/post-123"

Agent behavior:

• Use Sitepanda to scrape the page
• Read the extracted Markdown
• Summarize the main points in the response

Example 2

User request: "What does this documentation page say? https://example.com/docs"

Agent behavior:

• Fetch the page using Sitepanda
• Extract key sections
• Explain the content concisely