Autonomous AI agent networks resist Sybil attacks by enforcing proof-of-stake anchored in graph centrality metrics. At the dawn of each epoch, incoming agents submit cryptographic collateral through an ingress validation jetty. Identities are then mapped into a Merkle-based trust grove, where reputation is transitively bounded. This architecture forces an attacker to commit non-recoverable capital and compute per identity, making mass identity generation economically prohibitive.