Command principal
Who is submitting a command: a kind (agent, person, or service), a stable id, and the claims the caller was granted. Defined by ADR#0026 and carried into execution alongside the command itself, where an opt-in authorizer decides whether that principal may run it. An execution with no authorizer configured never asks; an execution with one refuses a command that carries no principal at all.
The principal is an input the runtime trusts rather than a claim it verifies. It is only as trustworthy as whatever produced it, which is why deriving one from a verified credential is a separate concern that lives at the edge.
Distinct from the channels principal, which is the identity behind an endpoint and governs message routing rather than command execution.
