All three tools lack input parameter schemas entirely (empty {} for all), violating the critical requirement that every tool must have a documented input schema. Descriptions are present but minimal (18-24 chars), below the 34-char baseline for p10. No parameter descriptions exist. Output schemas are undocumented. Tools are read-only and well-intentioned (safety-focused), but definition quality is severely deficient for LLM-driven agent use.

• Add explicit input schemas to all tools. Even if no parameters are required, declare `"properties": {}` with a clear description: 'This tool requires no parameters. Call it to retrieve current network health statistics.'
• Expand tool descriptions to 50 - 150 characters. Example: 'Retrieve network health statistics (uptime, latency, throughput) from the configured UniFi site. Call this to monitor overall network status.'
• Document output schemas in tool descriptions or via structured annotations. Example: 'Returns a dict with keys: uptime (%), latency (ms), throughput (Mbps), device_count (int).'
• Add limit and offset parameters to devices() and clients() to support pagination. Return a total_count field so agents can iterate through large result sets.
• Rename tools to verb_noun pattern: get_network_health, list_devices, list_clients. This helps LLMs parse intent from the name alone.