Four tools with basic schemas and descriptions, but significant gaps in parameter documentation and output schema clarity. Tool names follow verb_noun convention (list_, get_, refresh_, query), which is good. However, descriptions are minimal (10-50 chars), parameter descriptions are sparse or missing, and output schemas are not documented. The 'query' tool conflates two concerns (SQL generation vs execution) and lacks guidance on when to use each mode. No error recovery hints, no pagination support despite potential for large result sets, and no tool annotations (readOnlyHint, idempotentHint). Risk labels are present but not integrated into descriptions.

• Expand tool descriptions to 100-200 chars. Example: 'list_databases: Returns all configured PostgreSQL databases with connection status and schema cache state. Call this first to discover available targets before querying or refreshing schemas.'
• Document output schemas for each tool. E.g., list_databases returns [{database_id: string, name: string, status: 'connected'|'error', schema_cached: boolean}]. This enables LLM planning.
• Split 'query' into two tools: 'generate_sql' (takes natural language, returns SQL string) and 'execute_query' (takes SQL, returns results with pagination). Each has one responsibility.
• Add error recovery hints to descriptions. E.g., 'refresh_schema: If timeout occurs, the schema cache may be stale. Retry after 30s or call get_schema to use cached version.'
• Add limit and offset parameters to 'query' when output_mode='results'. Document max result size (e.g., 100 rows) and require pagination for larger datasets.

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

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