Three tools with basic verb-noun naming (current_weather, hourly_forecast, daily_forecast). All have short descriptions (20-40 chars) and minimal parameter documentation. Input schemas present but lack depth: location is a bare string with no format/length constraints; hours/days parameters have defaults but no min/max bounds. No output schema documented. Error handling converts exceptions to ValueError but provides no recovery guidance. Tools return JSON strings rather than structured objects, forcing LLM parsing. No tool annotations (readOnlyHint, idempotentHint). Descriptions are too brief to guide LLM selection or explain when to use each tool vs alternatives.

• Expand tool descriptions to 80-150 chars. Example: 'current_weather: Returns current temperature, humidity, wind speed, and conditions for a location. Use this for immediate weather info; use hourly_forecast for next 24 hours, daily_forecast for 7-day outlook.'
• Add parameter descriptions and constraints. location: 'City name (e.g. London, New York) or coordinates (lat,lon). Max 100 chars.' hours: 'Number of hours to forecast (1-168, default 24).' days: 'Number of days to forecast (1-16, default 7).'
• Document output schema in tool docstrings. Example: 'Returns {temperature: float (°C), humidity: int (%), wind_speed: float (m/s), description: string, timestamp: ISO8601}.'
• Add tool annotations: @mcp.tool(readOnlyHint=True) to mark all three as read-only (safe to retry). This signals to agents that these calls have no side effects.
• Improve error messages with recovery hints. Example: 'Location "xyz" not found. Try a major city name (London, Paris) or coordinates (51.5,-0.1). Call search_locations() if unsure.'

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

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