Team context layer

A git repo of what a data team knows (project history, data definitions and traps, house playbooks), written for people and AI. Analysts never browse it. Their AI assistant reads it at the start of a task and writes back what it learned at the end.

Every write is a pull request. Scripted checks run first, then an AI reviewer checks evidence: shared facts need a source, a method and a date. A scheduled steward hunts for stale entries and drift. A plugin on top carries the team’s methods into every analyst’s sessions.

I built it bottom to top on my own initiative, moved it into production, and still own every piece of it. It went from a notes page to a reviewed repo in four days, and to the whole data department in about two weeks.

The full story is in The data layer is three layers now.