Server has 5 tools with complete input schemas and descriptions. Naming follows verb_noun pattern (list_, get_, send_, execute_). Descriptions are adequate (100-250 chars) but lack actionable guidance on when to use each tool vs. alternatives. Parameter descriptions are present but sparse. No output schemas documented. Error handling exists but lacks recovery guidance. Security is strong (input validation, no secrets in params). Missing tool annotations (readOnlyHint/destructiveHint).

• Add outputSchema to each tool definition documenting the response structure. E.g., switchbot_list_devices returns {device_count: number, devices: [{deviceId, deviceName, deviceType, hubDeviceId}], infrared_remote_count: number, infrared_remotes: [{deviceId, deviceName, remoteType}]}.
• Add tool annotations: readOnlyHint=true for list_devices, get_status, list_scenes; destructiveHint=true for send_command, execute_scene. This signals to agents which operations are safe to retry.
• Define the 'parameter' field in switchbot_send_command with explicit type: oneOf [{type: 'string'}, {type: 'object'}] and document when each is used (e.g., 'String for simple params like "default", object for structured params like {position: 50}').
• Enhance error messages with recovery hints. E.g., catch ValueError for invalid device_id and return 'Invalid device_id. Call switchbot_list_devices to see available devices.'
• Document device type → command mapping in switchbot_send_command description. E.g., 'For Bot devices: turnOn, turnOff. For Curtain: setPosition (0-100). For Humidifier: setMode (auto, manual).' This prevents invalid command/device pairs.