sql-correctness

To validate Databricks SQL queries against Unity Catalog conventions, enforce 3-level catalog.schema.table naming patterns and reject unqualified table usage in agent traces. This ensures statements meet platform-specific execution requirements before running.

The correct way to execute Spark SQL on Databricks from an AI agent is mandating the mcp__databricks__execute_sql tool without Bash or notebook workarounds. This enforces safe execution and prevents prohibited CLI bypass behaviors.

To detect unsafe SQL interpolation in Databricks agent traces, evaluate execute_sql tool calls and SQL snippets against a correctness rubric that blocks risky string interpolation and unintended DROP operations before platform execution.

Yes, Unity Catalog requires fully qualified 3-level catalog.schema.table names for Databricks SQL validation. Enforcing this pattern rejects unqualified table usage and ensures queries meet platform-specific correctness requirements.

The limitations of using Bash workarounds for Databricks SQL execution include violating safety guardrails that mandate mcp__databricks__execute_sql. Disallowed Bash and notebook workarounds bypass Unity Catalog validation and platform-specific syntax checks.

To enforce modern DDL constructs in Spark SQL statements, apply a validation rubric that prefers CREATE OR REPLACE patterns, adds documentation via COMMENT ON, and updates table properties using SET TBLPROPERTIES before execution.