Scoring was not performed

• This repository is a CrewAI application, not an MCP server. If the goal is to expose scrape_as_markdown as an MCP tool, refactor to use the MCP SDK (Python: mcp library). Register the tool via server.tool() decorator or equivalent registration pattern.
• Expand scrape_as_markdown description to 100 - 150 chars explaining: 'Converts a web page (HTML) to cleaned markdown format. Use this to extract article text, documentation, or web content for analysis. Supports HTTP/HTTPS URLs. Returns markdown-formatted text with headings, lists, and links preserved.'
• Document the output schema explicitly: 'Returns {"markdown": string, "title": string (optional), "length": integer}. Markdown is guaranteed to be valid UTF-8 with no embedded HTML tags.'
• Add parameter constraints: 'url must be a valid HTTP or HTTPS URL (max 2048 chars). Timeout is 30 seconds, URLs taking longer return a timeout error with the partial content retrieved so far.'
• Implement error recovery guidance. Examples: 'If URL returns 403 Forbidden, the site may block automated scraping. Try a different source. If timeout occurs, the site may be slow, retry after 60 seconds or skip to next source.'
• Document timeout behavior, rate limits, and any retry logic. Agents need to know if they can call this tool in a loop or if there are per-domain request limits.
• Consider adding optional parameters: timeout_seconds (default 30), include_metadata (boolean, default true), max_markdown_length (limit output size). These give agents control over behavior.