ArticleLord has 6 tools with basic schemas and descriptions, but significant quality gaps. Tool naming is mostly verb-driven (open_webpage, close_popups, read_page_content, summarize_content, send_email, read_and_send), but read_and_send violates single-responsibility principle by combining three actions (open, summarize, send). Descriptions are present but brief (10-60 chars), below the 50-200 char LLM-optimized baseline. Parameter descriptions are minimal or absent. No output schemas documented. Error handling is not visible in the provided code. The read_and_send tool is particularly problematic, it should be split into composable pieces. No input validation guidance, no recovery hints, and no parameter constraints (enums, ranges) visible.

• Split read_and_send into three separate tools: open_webpage (already exists), summarize_content (already exists), send_email (already exists). Remove the composite tool and let agents compose them.
• Document output schemas for all tools. E.g., open_webpage should return {page_url, page_title, page_content_available}; summarize_content should return {summary, token_count, source_url}.
• Expand tool descriptions to 100-150 chars. Include WHEN to use, WHAT it returns, and any prerequisites. E.g., 'Summarize web content using Claude. Call after read_page_content to extract key points. Returns summary text and token count.'
• Add descriptions to all parameters. For summarize_content: 'title' → 'Optional title to include in the summary context'; 'max_tokens' → 'Maximum tokens for summary output (default 1024, range 100-4096)'.
• Add error handling guidance in tool descriptions. E.g., 'If open_webpage fails, check the URL format. If Playwright crashes, retry once. If timeout occurs after 30s, the site may be unreachable.'