Data liberation means getting public-interest data out of wherever it’s stuck — a removed page, a stack of PDFs, an undocumented portal — and into clean, documented, durable form. The lab ships it as two composable, MIT-licensed agent skills you can run inside an AI coding assistant like Claude Code, Copilot, or Gemini:

• data-liberation extracts, tidies, validates, and documents a messy source into a reproducible, citable dataset.
• data-project builds the right-sized project around it — documentation, pipelines, governance, and a publishable knowledge bundle.

Both climb in graded levels: start with a plain CSV, or go all the way to publication.

What’s new

The project template now lives inside the liberation skill (agent-driven scaffolding, no separate repo to pin), data-project ships Python and R stacks, and finished datasets archive to Harvard Dataverse with citable DOIs — the same engine behind the Colorado Environmental Data Hub. Still on the roadmap: the governance, responsible-data, and accessibility safeguards each skill defers until a project is ready to publish sensitive material.

When to reach for it

This is the engine behind the help desk’s everyday requests — “a website’s data disappeared”, “I have a bunch of PDFs”, and “combine data from different sources”. Installation and usage live in each repo’s README; browse our other tooling & code and tutorials & how-tos, or bring a dataset to the help desk.