Skip to main content

DuckDB Skills - Query Any Data Format from Claude Code

·158 words·1 min
Kostiantyn Lysenko
Author
Kostiantyn Lysenko

DuckDB reads CSV, JSON, Parquet, Avro, Excel, spatial formats, SQLite, and Jupyter notebooks. It connects to PostgreSQL, MySQL, and SQLite databases. It fetches files from S3, GCS, Azure, and HTTPS URLs. duckdb-skills opens all of these to Claude Code through a single plugin.

/plugin marketplace add duckdb/duckdb-skills
/plugin install duckdb-skills@duckdb-skills

The plugin provides six skills: attach-db, query, read-file, duckdb-docs, read-memories, and install-duckdb.

read-file detects the format, reads the schema, and previews the data:

/duckdb-skills:read-file data.parquet

query accepts plain English or raw SQL:

/duckdb-skills:query Average stars and issues by language from projects.csv?
language,avg_stars,avg_issues,projects
TypeScript,12600,77,2
Python,10133,70,3
Rust,8767,22,3
Go,6550,38,2

Ad-hoc queries run in a sandbox — DuckDB limits file access to the referenced files, blocks external connections, and locks the configuration. Queries against an attached database run with full access.

duckdb-docs searches DuckDB documentation offline. read-memories recalls context from previous Claude Code sessions.

Every format DuckDB supports, every provider it connects to — available in Claude Code with one install.

https://github.com/duckdb/duckdb-skills


comments powered by Disqus