ActionFence is a policy middleware library, not a primary MCP server. The 6 example tools have basic schemas and descriptions but significant gaps. Tool names are clear (search_flights, book_flight, check_status, bulk_booking, search_orders, delete_order), but descriptions are minimal (10-50 chars, well below the 194-char baseline). Parameters lack descriptions in most cases. No output schemas documented. Error handling and recovery guidance absent. The library itself focuses on policy enforcement (spend caps, approvals, schema drift) rather than tool quality. Example tools are illustrative but not production-grade.

• Expand tool descriptions to 100-200 chars. Example: 'Search available flights by route and date. Returns a list of flights with IDs, prices, and departure times. Use this to explore options before booking. Requires valid IATA airport codes (e.g., LAX, JFK) and a date in YYYY-MM-DD format.'
• Add descriptions to all parameters. For search_orders, document what 'query' accepts (free-text search, order ID, customer name?). For delete_order, clarify that orderId is the system ID and the operation is irreversible.
• Document output schemas for each tool. E.g., search_flights returns {flights: [{id, route, price, departure}], total_count}. This enables LLMs to chain tools correctly.
• Add error handling guidance. E.g., book_flight description: 'May fail if flight is sold out, passenger count exceeds availability, or amount is insufficient. Retry with a different flight or amount.'
• Implement confirmation pattern for delete_order. Add a dry_run parameter or require a separate confirm_delete_order tool to prevent accidental data loss.