The server provides 4 well-named database query tools with clear, action-verb names (list_, get_, describe_, query_). All tools have descriptions and input schemas with type definitions. However, several rubric violations reduce the score: (1) parameter descriptions are minimal, many lack guidance on expected format, constraints, or valid values; (2) output schemas are completely undocumented, no description of what fields or structure each tool returns; (3) error handling is present but not actionable, error messages are returned as plain text strings with no guidance on recovery steps; (4) the query_database tool accepts free-form SQL with no validation or sanitization, creating SQL injection risk and no user guidance on valid queries. The descriptions are adequate (50 - 100 chars) but could better explain WHEN to use each tool and HOW results differ. No tool annotations (readOnlyHint) despite all being read-only. Naming is strong (verb_noun convention), but parameter and output documentation falls short of production baseline.

• Document output schema for each tool. Example for list_tables: 'Returns {tables: string[]} containing sorted table names.' For query_database: 'Returns {columns: string[], rows: any[][]} where rows is an array of arrays matching column order.'
• Add constraint and format guidance to parameter descriptions. For table_name: 'Name of the table to describe (alphanumeric, no spaces; obtained from list_tables).' For sql: 'SELECT query only. No INSERT, UPDATE, DELETE, or DDL. Use standard SQL syntax.'
• Implement input validation for query_database. Reject non-SELECT queries with actionable error: 'Only SELECT queries are allowed. Attempted: UPDATE... Did you mean: SELECT * FROM ...'
• Add recovery guidance to all error responses. Pattern: 'Error type: [description]. Next step: [suggested action]. Example: "Table not found. Call list_tables() to see available tables."'
• Add tool annotations to tool definitions. Use readOnlyHint: true for all 4 tools to signal to agents they are safe to retry.
• Add pagination parameters (limit, offset) to list_tables and query_database. Cap results at 50 rows by default, return total_count and has_more in response.
• Distinguish between list_tables and get_db_schema in descriptions. Example: 'Call list_tables() for a quick table inventory. Call get_db_schema() for detailed column schemas (types, constraints, nullability).'
• Add per-tool usage guidance in descriptions. Example for query_database: 'Use this to read data. For schema inspection, prefer describe_table() or get_db_schema().'

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

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