What problem does it solve? Building a Malloy semantic model without understanding the underlying data leads to broken joins, wrong grains, and invalid measures. This Skill performs silent, query-grounded data discovery so every modeling decision is backed by actual data rather than schema guesses. ## Core Features & Use Cases - Grounded Profiling: Runs discovery queries through Malloy's execute_query tool to check NULL rates, cardinalities, distributions, and duplicate keys. - Join and Grain Validation: Verifies join cardinality, key value compatibility, and sibling tables at the same grain before proposing relationships. - Prior Art Detection: Detects LookML files, dbt projects, metadata files, and KPI docs, then reconciles their claims against queried data. - Use Case: Before writing a Malloy model for a new analytics database, run this discovery step to record table roles, data quality issues, and validated join paths in modeling-notes.md for the scope proposal. ## Quick Start Use the malloy-discover skill to silently profile the connected data sources and record findings in modeling-notes.md before proposing a model scope.
