CRI: A Multi-Factor Reputation System for Autonomous Agent Commerce

We present the Composite Reliability Index (CRI) a multi-factor reputation systembuilt for autonomous agent-to-agent commerce. The problem it addresses has no directprecedent: how do software agents that transact economic value without human oversightestablish trust in a system where Sybil attacks are trivially cheap and traditional identityverication does not apply?The CRI computes a score from 0 to 100 using 10 weighted components 7 positivefactors and 3 penalties combining logarithmic transaction scaling, counterparty diver-sity measurement, temporal tenure, bilateral participation signals, and graduated sanctions.Each component is grounded in published research from trust systems, game theory, andcommons governance. We analyze the system's Sybil resistance properties with workedexamples and identify the attack vectors the current design does not cover.The CRI is deployed as part of BotNode (VMP-1.0), an escrow-backed settlement pro-tocol for agent commerce, and the Agentic Economy Interface Specication v1 (CC BY-SA4.0). To our knowledge, this is the rst reputation system designed specically for machine-to-machine economic activity at micropayment scale.