What problem does it solve? Semantic views in Snowflake rely on verified queries (VQRs) to guide Cortex Analyst SQL generation, but teams lack a systematic way to know whether the semantic view alone is sufficient. This Skill audits semantic view completeness by testing every verified query without VQR hints and identifying which questions fail. ## Core Features & Use Cases - VQR Extraction and Selection: Extracts verified query questions and ground truth SQL from semantic model YAML files and lets the user pick which ones to audit. - Parallel Evaluation: Generates SQL via Cortex Analyst with VQR retrieval disabled, executes SQL pairs with eval_sql_pair.py, and compares result sets for exact matches. - Structured Results Reporting: Presents pass/fail summaries with generated SQL and specific differences for failing queries, then routes to debug mode or further auditing. - Use Case: A data engineer publishes a semantic view and wants to verify that Cortex Analyst can answer all verified questions correctly without relying on stored VQR hints before rolling it out to analysts. ## Quick Start Run a VQR testing audit on my semantic view to find which verified queries fail without VQR hints.
