Server has 4 tools with basic schemas and descriptions, but significant quality gaps. Tool names lack clear action verbs (natural_language_query is vague; execute_sql is generic). Descriptions are present but brief (avg ~80 chars, below 194-char baseline). Parameter descriptions are minimal or absent. No output schemas documented. No error handling guidance. No tool annotations. Security concerns: execute_sql accepts arbitrary SQL with minimal validation. Missing pagination, result limits, and structured output guidance.

• Rename tools to be more specific: 'convert_nl_to_sql' (returns SQL only), 'execute_sql_query' (executes), 'describe_database_schema', 'validate_sql_syntax'. Each should have one clear responsibility.
• Add output schemas to all tools. Document what fields are returned, their types, and what the LLM should do with them. E.g., execute_sql returns {rows: [{col1: string, col2: number}], row_count: number, execution_time_ms: number}.
• Expand parameter descriptions with format/constraint guidance. E.g., 'query: Natural language question about the database (e.g., "How many users signed up last month?"). Avoid complex joins; use simple, direct questions.'
• Implement parameterized queries and input sanitization for execute_sql. Never interpolate user input into SQL strings. Use psycopg2 parameter binding (e.g., cursor.execute('SELECT * FROM users WHERE id = %s', (user_id,))).
• Add error classification and recovery guidance. Return {error: string, error_type: 'retryable'|'user_fixable'|'fatal', suggestion: string}. E.g., 'Table not found. Available tables: users, orders, products. Did you mean one of these?'

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

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