snowflake-semanticview

This skill guides the assistant through creating, altering, and validating Snowflake semantic views using the official Snowflake CLI (snow). It solves the problem of building reliable semantic-layer definitions by enforcing a disciplined, validation-first workflow: rather than producing untested DDL, the assistant drafts CREATE/ALTER SEMANTIC VIEW statements, validates them against Snowflake via the CLI using a temporary name, iterates until they pass, and only then applies the final definition. It also provides one-time setup guidance for installing the Snowflake CLI and configuring a connection via snow connection add, pointing to official Snowflake documentation.

The workflow is thorough. It has the assistant confirm the target database, schema, role, warehouse, and view name, verify a star-schema model, and draft DDL from the official syntax. It requires populating synonyms and comments for every dimension, fact, and metric, preferring existing Snowflake column comments as the source and asking permission before inventing any. It uses read-only SELECT statements with DISTINCT and a LIMIT of at most 1000 rows to discover relationships and column types, validates DDL under a temporary name (for example a __tmp_validate suffix) to avoid clobbering the real view, runs a sample SEMANTIC_VIEW query to confirm behavior, and cleans up temporary objects afterward.

The target users are data engineers and analytics engineers building Snowflake semantic layers or metric definitions. Credential handling is delegated to the official Snowflake CLI connection configuration rather than being managed by the skill, and the emphasis on temporary validation names, LIMIT caps, and asking before creating comments reflects careful, low-risk data practices.