Four read-only tools with minimal descriptions and no parameter descriptions. All tools follow a consistent verb_noun naming pattern (get_*) and accept a single supplier_id parameter with basic type info. However, descriptions are generic (10-30 chars), parameter descriptions are absent, output schemas are undocumented, and error handling is minimal. No tool annotations, no pagination, no structured output guidance. The server is functionally simple but lacks the documentation depth required for LLM-driven agent selection and error recovery.

• Expand tool descriptions to 80-150 chars. Example: 'get_supplier_risk: Retrieve the current risk assessment for a supplier (low, medium, high, critical). Use this to evaluate procurement risk before approving orders. Returns risk level and last assessment date.'
• Add parameter descriptions. supplier_id should be: 'The unique supplier identifier (string, e.g., SUPP-001 or vendor-acme-corp). Obtain from search_suppliers() or procurement records.'
• Document output schemas in tool descriptions or via a separate schema registry. Example: 'Returns {risk_level: string (low|medium|high|critical), supplier_id: string, assessed_at: ISO8601 timestamp}'.
• Enhance error messages with recovery guidance. Instead of 'unknown supplier: SUPP-999', return 'Supplier SUPP-999 not found. Available suppliers: SUPP-001, SUPP-002, SUPP-003. Call search_suppliers(name_pattern) to find by name.'
• Add tool annotations via FastMCP metadata. Mark all four tools with readOnlyHint=true and idempotentHint=true to signal to agents that these are safe to retry and do not mutate state.

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

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