Server has 5 tools with complete schemas and descriptions, but quality is inconsistent. Naming follows verb_noun convention (list-, create-, get-, update-). Descriptions are present but generic (avg ~50 chars, below 194-char baseline). Parameters have types and descriptions, but lack enums for constrained values (e.g., status field accepts free-form strings). Output schemas are not documented. Error handling returns generic messages without recovery guidance. No tool annotations (readOnlyHint, destructiveHint). create-order and update-order-status are write operations but lack confirmation/dry-run patterns.

• Add formal output schemas for all tools. Document return types: list-products returns {data: Product[]}, create-order returns {data: {order_id, status, message}}. This enables LLM planning.
• Convert status parameters to enums. Replace z.string() with z.enum(['pending_confirmation', 'delivered', 'cancelled', 'refunded']) in list-orders and update-order-status.
• Expand tool descriptions to 100 - 200 chars. Include WHEN to use, prerequisites, and dependencies. E.g., 'Admin function to update order status. Requires order_id from get-order or list-orders. Valid statuses: pending_confirmation, delivered, cancelled.'
• Add tool annotations: mark list-products, list-orders, get-order as readOnlyHint=true; mark create-order, update-order-status as destructiveHint=true. This signals to agents which tools have side effects.
• Implement actionable error messages. When product_id is invalid, return: 'Invalid product_id: PZ999. Available products: PZ001 (Choco Blast), PZ002 (Peanut Butter Power). Call list-products to see all options.' This enables self-correction.

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry