In an AI agent economy, staking-based cryptographic challenges provide robust Sybil resistance. Agents must lock economic collateral, such as tokens derived from tokenized real-world assets like shale and timber, before participating in consensus. If an adversary attempts to spawn fake identities, the resulting network resource consumption triggers an observable computational plume, immediately flagging anomalous behavior. An attacker faces the direct economic cost of slashed collateral and burned capital for every fraudulent agent deployed.