Agent skill · personamanagmentlayer
analytical-databases-expert
Design and query columnar analytical stores: DuckDB, ClickHouse and cloud warehouses, including file formats, partitioning, sort keys and cost control. Use when the user mentions DuckDB, ClickHouse, Parquet, columnar storage, OLAP, a data warehouse or lakehouse, analytical queries over large tables, or when the task involves aggregating billions of rows, choosing between a warehouse and an embedded engine, or making a reporting query fast enough.
What it needs
About 6k tokens when loaded.
What this skill does
Analytical Databases Expert Columnar engines answer aggregate questions over large tables. They are not faster row stores — they trade single-row access for the ability to scan one column of a billion rows without touching the rest. Core Concepts Why Columnar Wins Storing values of one column contiguously means a query reads only the columns it names, and values of one type compress together far better than mixed rows. SELECT avg(amount) FROM events touches one column; the row store reads every byte of every row. The corollary is the cost: fetching one complete row means reassembling it from every column, and single-row updates are expensive or unsupported. Use a row store for transactions and a columnar store for analysis. Trying to make one system do both well is the usual architectural mistake. The Data Is Often the Format With Parquet, storage and engine are separable. Files on object storage, queried by whatever engine suits: DuckDB locally, Spark for scale, the warehouse for governed access. Choosing the format well matters more than choosing the engine. Selection Is About Deployment, Not Speed Engine Fits Cost --------------------- ------------------------------------------------------------------- ------------------------------------------------------- DuckDB Embedded analytics, notebooks, CI, single-node up to hundreds of GB Single process; no concurrent writers ClickHouse High-ingest real-time analytics, dashboards over billions of rows Operationally demanding; eventual-consistency semantics Cloud warehouse Governed shared access, elastic scale, SQL for everyone Cost grows with carelessness; vendor coupling Spark / lakehouse Petabyte batch, heavy transformation, ML feature pipelines Latency and complexity DuckDB is the default worth trying first: most "big data" is under a terabyte, and a single node handles it. DuckDB Runs in-process. No server, no cluster, and it queries Parquet, CSV and JSON on local disk or object storage directly. …
How to use it
Reference it in AdaL, Claude Code, Cursor or any coding agent — nothing to install:
@skills personamanagmentlayer/analytical-databases-expert