Methasit-Pun Data Modeling & Transformation (Router)
This is a router skill. It groups the four skills that deal with structuring and transforming data after it lands in the warehouse or a processing job. Diagnose which sub-area(s) the task touches, then invoke the matching sub-skill(s) with the Skill tool.
How to route
| If the task is about… | Invoke sub-skill |
|---|---|
| Modeling business entities: star/snowflake schema, one-big-table, slowly changing dimensions, grain definition, surrogate keys, normalization tradeoffs | schema-design |
| Writing/reviewing analytical SQL: window functions, CTEs, ranking, running totals, dedup, query optimization, anti-patterns (BigQuery/Snowflake/Redshift/DuckDB) | sql-patterns |
| dbt work: model layers (staging/intermediate/marts), refs, sources, tests, macros, materializations, incremental strategies | dbt-patterns |
| Python transforms: pandas/Polars/PySpark idioms, chunked reads, memory-safe transforms, vectorization, performance | python-data-patterns |
Routing rules
- Start with
schema-designwhen the table doesn't exist yet — decide grain and layout before writing SQL against it. A bad schema is expensive to fix later. - SQL running inside dbt pulls in both
sql-patterns(the query itself) anddbt-patterns(materialization, refs, tests). - Transformation in Python vs. SQL undecided? Warehouse-native transforms →
sql-patterns/dbt-patterns; external/large-file or ML-adjacent processing →python-data-patterns. - Row-by-row DataFrame iteration or memory errors →
python-data-patternsimmediately. - Invoke via the Skill tool by name, e.g.
Skill(skill="dbt-patterns"). Combine outputs; don't paraphrase from memory.
Related groups
- Getting data into the warehouse first → [[data-pipelines]]
- Testing/validating the models you build → [[data-reliability]]
- Warehouse cost of these queries → [[cloud-data-infra]]