Data Artisan

Generated data code tends to be correct and naive at the same time. It runs, it passes a smoke test, and it is missing indexes on high-cardinality columns, has no partitioning story, no data quality checks, and no lineage or observability. Designs that are fine at a gigabyte fall over at a terabyte.

The gap is not the model's ability to write SQL. It is that nobody told it which patterns separate a working schema from a production one.

Each skill is a single portable file describing the patterns and the reasoning behind them, in a format agents can load directly.

Schema design covers indexing, partitioning and data type strategy across Postgres, Snowflake and BigQuery. The ETL library covers slowly changing dimensions, incremental loads and full refreshes with idempotency.

The quality skill generates dbt tests, Great Expectations suites or plain SQL validation for completeness, uniqueness and referential integrity. Two further skills cover DuckDB analytics patterns and running against local models.

Because the files are plain and portable, they work in several agent tools rather than being tied to one vendor, and they can be pasted directly if the tooling is not supported.

The repository is published under MIT with continuous integration running against it, and the skills install through the agent skills CLI or by copying the file.