Delegation Chains

A delegation chain is a sequence of cryptographically-signed certificates that links an AI agent back to a human principal.

Chain Structure

Human Principal ──[cert 1]──> Deployer Agent ──[cert 2]──> Worker Agent
     depth: 0                    depth: 1                    depth: 2

Every chain must terminate at a human root. Chains without a human root are invalid — the agent operates at SHADOW tier.

Certificate Fields

Each DelegationCert contains: delegator public key, delegate public key, action scope, resource scope, territory binding, issuance time, and optional expiration.

Verification

DelegationChain.verify() walks the chain from agent to root, checking that every certificate is valid, not expired, and that the root is a human identity (not another agent).

Scope Narrowing

Each delegation can only narrow scope, never widen it. If a human delegates ["search", "code"] to Agent A, Agent A cannot delegate ["search", "code", "payment"] to Agent B.