The rules file grew up and got a registry

Every AI tool grew its own dialect of the same idea: a file that teaches the model your conventions. They live per-tool, per-laptop, per-repo, and they drift.

Hoody agents read skills: files in the container, managed by a registry with an API. Write the deployment checklist once, then PUT it into every project container that needs it. Import a community skill from the hub instead of writing your own. When a skill changes, it changes the way a file changes: versioned, diffable, deployed like code.

The knowledge stopped being tool configuration and became infrastructure.