18 tools with consistent naming (verb_noun pattern), clear descriptions (avg 85 chars), and proper JSON schemas with typed parameters. All tools are read-only with correct annotations. However, output schemas are not documented, LLMs cannot predict response structure. Parameter descriptions are minimal (often single phrases). Error handling returns generic dicts without recovery guidance. No pagination support despite tools returning lists. Missing parameter constraints (enums, ranges) for filter parameters like 'action', 'protocol', 'interface'.

• Document output schemas for all tools. Example: 'Returns list of {ip: string, mac: string, hostname: string, interface: string}' for get_arp_table. Use JSON Schema in tool definition or explicit field descriptions in docstring.
• Add enum constraints to filter parameters. Change 'action' parameter from free-form string to enum: ['pass', 'block', 'reject']. Add enums for 'protocol' (tcp, udp, icmp, etc.) and 'interface' (list valid interface names from list_firewall_interfaces).
• Implement cursor-based pagination. Add 'offset' and 'cursor' parameters to list-returning tools. Return 'total_count' and 'next_cursor' in responses. Document in tool descriptions: 'Returns up to 200 entries; use next_cursor to fetch additional results.'
• Enhance error handling with recovery guidance. Instead of {'error_type': 'ConnectionError'}, return {'error': 'Failed to connect to pfSense WebGUI', 'retryable': true, 'suggestion': 'Check base_url and credentials via check_webgui_login()'}. Categorize errors as retryable/user-fixable/fatal.
• Expand parameter descriptions with format and range constraints. Example: 'ip_address: IPv4 address to filter by (e.g. 192.168.1.100). Accepts CIDR notation (e.g. 192.168.1.0/24). Optional.' Include valid ranges for 'limit' (1-1000).