I found that AI coding tools work well for one person, but teams quickly get drift: one person’s Cursor rules, another person’s Claude context, old decisions in Slack, quality standards in someone’s head.
I started using a simple pattern: keep outcomes, customers, decisions and quality bar in the repo, then let each AI tool read/adapt from that same source.
I packaged the pattern as an open-source CLI: npx create-team-foundry.