Three tools with complete schemas and descriptions, but descriptions are brief (10-20 chars baseline vs 194 char production average). All tools have proper input schemas with types and descriptions. Error handling present but minimal, calculator handles division by zero, datetime validates timezone, fetch-url truncates large responses. No output schemas documented. No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Naming is clear and verb-based (calculator, datetime, fetch-url), but descriptions lack context on WHEN to use each tool or what downstream tools might need. fetch-url lacks security guidance on URL validation and response sanitization.

• Expand tool descriptions to 100-200 characters. Include WHEN to use (e.g., 'Use calculator for precise arithmetic when LLM math may hallucinate'), WHAT it returns (e.g., 'Returns formatted equation result as text'), and any prerequisites (e.g., 'Requires valid IANA timezone name').
• Document output schemas for all tools. Define response structure: calculator returns {type: 'text', text: string}; datetime returns {type: 'text', text: string}; fetch-url returns {type: 'text', text: string, truncated: boolean, original_length: number}.
• Add tool annotations via server.tool() metadata: readOnlyHint=true for all three tools, idempotentHint=true for all three. This enables agents to reason about safety and retry behavior.
• Enhance fetch-url description with security context: 'Fetches HTTP GET content. Validates URL format. Truncates responses >10KB. Do not use with untrusted URLs or sensitive endpoints.' Add timeout_ms parameter description: 'Timeout in milliseconds (1-60000, default 5000). Prevents hanging on slow/unresponsive servers.'
• Improve error messages. For fetch-url truncation, return {type: 'text', text: '...', truncated: true, original_length: 15000} so agents know data was cut. For datetime invalid timezone, suggest valid alternatives: 'Invalid timezone "Foo". Try: UTC, America/New_York, Asia/Tokyo, Europe/London'.