What problem does it solve?
This Skill provides an expert-level, modular assistant for running rigorous DuckDB performance work: it orchestrates TPCH benchmarking, interprets execution plans for query optimization, and explains DuckDB internals so engineers can diagnose and fix performance issues quickly and reproducibly.
Core Features & Use Cases
- Coordinated Subagents: Delegates tasks to tpch-benchmarker, duckdb-optimizer, and duckdb-internals for focused expertise.
- TPCH Benchmarking: Generates TPCH data, runs strict 1 cold + N warm run benchmarks, logs everything, and produces DUCKDB_STATS.md and DUCKDB_ANALYSIS.md.
- Query Optimization: Interprets EXPLAIN / EXPLAIN ANALYZE output, identifies bottlenecks, and suggests targeted rewrites or configuration changes.
- Internals & Configuration: Explains columnar storage, vectorized execution, memory limits, threading, and trade-offs for performance tuning.
- Helper Script: Includes a reproducible duckdb_benchmark.sh script to enforce threads=1 and memory_limit='1GB' and to collect timing statistics.
Quick Start
Run a TPCH benchmark for query 5 on tpch.duckdb with 5 warm runs and save the results to DUCKDB_STATS.md and DUCKDB_ANALYSIS.md.