Principal Chain
The chain of on-whose-behalf an action runs: user to agent to sub-agent, every action resolving back to a human principal.
A principal chain answers one question for every action: who is this ultimately being done for? It links the human principal to the agent, and the agent to any sub-agents it spawns.
Why it matters
When agents spawn agents, accountability evaporates fast. If you cannot trace an action back to a human principal, you cannot enforce permissions, attribute responsibility, or revoke access cleanly. The chain keeps a human at the root of everything that happens.
In practice
Each action records its principal and the delegation path that produced it. A sub-agent inherits a narrowed slice of its parent’s authority, never more. You can ask, for any side effect, which human it was ultimately acting for.
Where it shows up in my work
This is core to governable-ai.com. Delegated machine action only stays safe when every action resolves to a principal with bounded, inheritable authority.