Single tool with minimal definition quality. Tool name 'extract_web_content' is appropriately verb-noun structured. Description is present but generic (10 words, 57 chars), lacks context on when to use, what it returns, or prerequisites. Input schema visible with 'url' parameter typed as string with basic description, but no constraints (format, length, allowed schemes beyond http/https check). No output schema documented, LLM cannot predict return structure. No error handling guidance; exceptions return raw error strings. No pagination, no structured output, no field descriptions beyond the single parameter. Lacks idempotency guarantees, rate-limit guidance, or timeout documentation.

• Expand tool description to 100-150 chars: 'Extract visible text content from a web page. Returns cleaned, line-separated text with scripts and styles removed. Use when you need to read or summarize web content. Supports http:// and https:// URLs only.'
• Document output schema: 'Returns a string containing the extracted text, with one paragraph per line. Empty lines removed. Returns error message string if extraction fails.'
• Add input constraints to 'url' parameter: type=string, pattern='^https?://', minLength=10, maxLength=2048, description='The URL of the web page (must start with http:// or https://)'
• Implement structured error responses with recovery hints: 'Invalid URL format. Must start with http:// or https://. Example: https://example.com' or 'Connection timeout after 30s. URL may be unreachable or very slow. Try a different URL or check connectivity.'
• Add timeout and idempotency documentation to description: 'Requests timeout after 30 seconds. This tool is read-only and safe to retry on failure.'

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry