This server has two tools with visible schemas and descriptions, but significant quality gaps prevent a higher score. The search_knowledge tool has a reasonable schema with enum constraints for intent, but the create_ticket tool uses z.any() for its primary input, which violates schema discipline entirely. Descriptions are present but generic and lack detail about when to use tools or what they return. Parameter descriptions are minimal. The 'intent' parameter appears in both tools but its enum is dynamically loaded from config (INTENT_NAMES), while this is flexible, it's not discoverable from static analysis of the schema alone. Error handling exists but provides minimal guidance to the LLM about recovery paths. The server uses Zod for validation (good practice) but does not leverage outputSchema fully, the search_knowledge tool returns structuredContent but most tools do not document their output structure clearly.

• Replace z.any() in create_ticket with a proper Zod schema matching the ticketSchema resource. Define the ticket structure explicitly so LLMs and validators can enforce it. Example: z.object({ title: z.string(), description: z.string(), priority: z.enum(['low', 'medium', 'high']) })
• Expand tool descriptions to 50 - 150 characters following pattern:tool-description. For search_knowledge: 'Search the knowledge base for documents matching a query. Returns up to k results ranked by relevance. Use this to find information before answering user questions.' For create_ticket: 'Create a support ticket from user query details. This is a destructive operation, confirm details before calling. Returns ticket ID and confirmation message.'
• Add descriptions to all parameters. Example for 'intent': 'The user's intent category (e.g., billing, technical, general). Determines which tools and resources apply to this query. Retrieved from the intents resource.' Do not use 'from config' in descriptions, LLMs cannot read config.
• Document the output schema for search_knowledge explicitly in the tool registration. Example: outputSchema: { results: z.array(z.object({ sourceId: z.string(), text: z.string(), score: z.number() })), count: z.number() }
• Improve error messages to guide recovery. For search_knowledge: 'No results found for query "{query}". Try: (1) search with simpler keywords, (2) check the knowledge resource for available documents, (3) try a different intent.' For create_ticket: 'Failed to create ticket: {error details}. Verify the ticket structure matches ticket_schema resource before retrying.'