etl

Run `npx skills add owid/etl --all -g -y` in your terminal to install all skills in this repository globally.

It moves external datasets through snapshot, meadow, garden, and grapher stages so raw source files become clean, documented, chart-ready data. The skills automate each stage, from creating snapshots to publishing updates.

The update-dataset skill runs the whole refresh end to end: bumping versions, rebuilding steps, checking metadata, and preparing the PR. It pauses only when a decision genuinely needs a human.

Yes. The adversarial-data-review skill verifies metadata claims against the producer's own documentation and cross-checks suspicious values against independent sources, catching unit errors and wrong-year values.

No. Skills like create-dataset are designed for non-experts: provide a data file or link, and the agent infers the structure, builds the pipeline, and asks you to review the result.