Mirror
A mirror is a runnable twin of the backend an AI agent talks to, built from its production traces.
What it is
A mirror holds the traces an agent produced, the world model reconstructed from them, and the builds that turn that model into a twin you can call.
Each build is a version; one is active, one may be pinned.
How it is used
Agents call the twin through the tool plane instead of production, so a change can be tried without touching real systems.