What problem does it solve? Bringing an existing data model from tools like dbt, LookML, or Snowflake into a markdown-based semantic layer is error-prone when source metadata is trusted blindly. This Skill imports external schemas into the vault while verifying every claim against real warehouse data before writing anything. ## Core Features & Use Cases - Preset-driven imports: Loads source-specific import presets from utils/imports/{source}.md that define how to connect, what to extract, and how to map source concepts to vault types. - Query-verified drafting: Runs verification queries against the actual warehouse to confirm grain, primary key uniqueness, column contents, join cardinality, and metric expressions before drafting any file. - Conflict-safe merging: Compares source metadata against existing vault files and surfaces disagreements on SQL, cardinality, aggregation, and descriptions for explicit user decision instead of silent overwrites. - Use Case: A team adopting dbt wants their models reflected in the Metalayer vault. The agent loads the dbt preset, extracts models and relationships, verifies each against Snowflake, drafts view, field, relation, and lore files, and presents them for approval before writing. ## Quick Start Tell the agent to update from your source, for example: update from dbt and import my models into the vault.
