The server provides 8 tools across two packages (agents and developers). Most tools have descriptions and some parameter documentation, but several critical gaps exist. The agent tools (get_data_products, get_data_product_details, query) have good, LLM-optimized descriptions explaining WHAT they do and WHEN to use them. However, input schemas are incomplete or missing validation details. The developer tools (execute_ddl, run_sql_transaction, monitor_data_freshness, object_freshness_diagnostics, search_documentation) lack visible schemas and detailed parameter documentation. Error handling guidance is absent across all tools. Security considerations (SQL injection, DDL injection) are not documented.

• Add complete input schemas for all 5 developer tools. Define parameter types (string, number, boolean), required vs optional, constraints (enums, patterns, min/max), and descriptions for each parameter.
• Document output schemas for all 8 tools. Specify what fields the response contains, their types, and example values. This enables LLMs to plan downstream actions and extract data correctly.
• Expand descriptions for developer tools to follow the pattern: WHAT does it do, WHEN to use it, WHAT does it return. E.g., execute_ddl: 'Execute SQL DDL statements (CREATE, ALTER, DROP) to define or modify database schema. Use when you need to create new tables, views, or materialized views. Returns success/failure with any errors or schema changes made.'
• Add error handling guidance to query, execute_ddl, and run_sql_transaction. Document common errors (syntax error, permission denied, cluster not found) and suggest recovery steps. E.g., 'If cluster not found, call get_data_products() to see available clusters.'
• Implement and document input validation and SQL injection safeguards. Use parameterized queries internally; document in tool descriptions that placeholders or prepared statements are recommended. Add security audit trail notes (who called what, when, with what parameters).
• Explicitly link get_data_product_details response to query() parameters. State in get_data_product_details description: 'The cluster field returned here must be passed to the query() tool's cluster parameter.'
• Add pagination support (limit, offset, cursor) to get_data_products and query tools with documented limits. E.g., 'Returns up to 100 data products; use cursor for pagination.' Cap results at a reasonable default (20-50) to prevent context window exhaustion.

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

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