Server provides 6 tools with complete input schemas and basic descriptions. However, descriptions are generic and lack actionable context for LLM selection. Parameter descriptions are minimal (e.g., 'Name of the connection to use (optional, uses default if not specified)' repeated across tools). No output schemas documented. No error handling guidance. Tool names follow verb_noun convention but descriptions do not explain WHEN to use each tool or what distinguishes similar tools (e.g., list-tables vs get-table-schema). Missing parameter constraints (e.g., max_rows has no bounds). No guidance on SQL injection prevention for execute-query.

• Document output schemas for all tools. Specify that list-connections returns {connections: [{name, status, driver}], default_connection: string}. Specify that execute-query returns {rows: [], columns: [], row_count: int}. This enables LLMs to chain tools and extract required fields.
• Add bounds to max_rows: 'Maximum number of rows to return (optional, default 100, range 1-10000). Prevents memory exhaustion and timeouts.'
• Enhance execute-query description: 'Execute a SQL SELECT query. Use parameterized queries to prevent SQL injection. Returns rows as JSON array. Supports SELECT only; INSERT/UPDATE/DELETE blocked for safety.'
• Replace generic connection_name descriptions with: 'Name of the configured ODBC connection (optional). If omitted, uses the default connection from config. Call list-connections first to see available options.'
• Add error handling guidance to tool descriptions: 'If connection fails, verify the connection name with list-connections. If SQL syntax error occurs, check query format. If permission denied, verify user credentials in config.'

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

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