duckdb-skills plugin
DuckDB plugin for Claude Code: SQL over CSV, Parquet and databases, S3 file reading, and search across docs and session logs
- 616 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The skills run the DuckDB CLI in the shell and install extensions, including community ones.
- `read-memories` reads past session logs and `state.sql` can hold cloud storage secrets.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/duckdb-skillsDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The skills run the DuckDB CLI in the shell and install extensions, including community ones.
- `read-memories` reads past session logs and `state.sql` can hold cloud storage secrets.
This is third-party code. Review the repository files before installing.
What it does
duckdb-skills is a plugin from the DuckDB team for Claude Code with a set of skills for data exploration. The `read-file` skill opens CSV, JSON, Parquet, Avro, Excel, spatial files, SQLite and Jupyter notebooks, local or from S3, GCS, Azure or over HTTPS, while `query` runs SQL or a natural-language question against an attached database or directly against files. `attach-db` attaches a DuckDB database file, inspects the schema and writes the state to `state.sql` so the other skills can restore the session. `duckdb-docs` searches the DuckDB and DuckLake documentation and blog, `read-memories` searches the logs of past Claude Code sessions, and `install-duckdb` installs extensions. The repository also has `convert-file`, `s3-explore` and `spatial`. Skills are invoked with a command like `/duckdb-skills:<name>`.
Who it is for. Analysts and data engineers who ask Claude Code to explore CSV, Parquet or a DuckDB database without a separate BI tool.
Good fit when
- A folder holds a Parquet export of hundreds of megabytes and its schema and row count are unknown.
- Data on S3 or Cloudflare R2 should be previewed without downloading the files.
- An Excel table is converted to Parquet for further work.
- A decision from two weeks ago is lost and `read-memories` looks for it in past session logs.
Not a fit when
- Data sits in a production DBMS with transactions: the plugin targets files and local DuckDB databases.
- The agent runs on Windows, where support is still incomplete.
- The DuckDB CLI cannot be installed and the skills call it from the shell.
Example request
Open sales.parquet, show the schema and row count, then compute the five largest customers by revenue.Limitations
The plugin is tested on macOS and Linux, and on Windows some shell commands and paths may not work. The DuckDB CLI is required: if it is missing, the skills offer to install it through `install-duckdb`. The `state.sql` file sits in the project or the home directory and can hold cloud storage secrets, so it should not be committed. The `read-memories` skill reads the logs of past Claude Code sessions in full. The last commit in the repository is from April 2026. The plugin is not yet in the official Anthropic marketplace and installs from GitHub.
How to disable. Run `/plugin uninstall duckdb-skills@duckdb-skills` and, if you want, `/plugin marketplace remove duckdb-skills`. Delete the `state.sql` files by hand.
FAQ
How do I install duckdb-skills in Claude Code?
Run `/plugin marketplace add duckdb/duckdb-skills`, then `/plugin install duckdb-skills@duckdb-skills`. The skills appear in new sessions as `/duckdb-skills:<name>`.
How do I run a query against a file?
Call `/duckdb-skills:query FROM 'exports.csv' WHERE amount > 100`. Instead of SQL you can ask a natural-language question and the skill writes the query.
Where is session state stored?
In a `state.sql` file with ATTACH, USE and LOAD statements, secrets and macros. The skill asks where to keep it: in the project `.duckdb-skills/` or in `~/.duckdb-skills/<project>/`.
How do I update the plugin?
First `/plugin marketplace update duckdb-skills`, then `/plugin update duckdb-skills@duckdb-skills`.
Which formats does read-file handle?
CSV, JSON, Parquet, Avro, Excel, spatial files, SQLite and Jupyter notebooks, from disk and from S3, GCS, Azure or over HTTPS. The format is detected by file extension.
Discussion
No questions or reviews yet.
Related
PluginsA temporary DuckDB sandbox: an agent loads a CSV and queries the data over MCP or REST without touching the production database
dbt Agent Skills
dbt-agent-skills
Official dbt Labs skills: the agent understands dbt projects, runs workflows and helps with migration
Three Claude Code plugins: queries to OpenData public datasets, OpenChart charts and tables, and SVG graphics.
DBTrail reads the MySQL binlog, keeps a Parquet copy for DuckDB queries and gives Claude Desktop read-only MCP access
The MCP server of the Duckle platform: an agent builds, checks and runs DuckDB ETL pipelines on your machine
Around 200 skills for Claude Code on bioinformatics, chemistry and biostatistics, from RNA-seq analysis to molecular docking
README in short
The plugin is built around a shared state file: skills extend it with databases, secrets and macros and the others pick them up without reconfiguration. You can try the plugin without installing by running `claude --plugin-dir .` from a clone of the repository.