AgentReputation: Decentralized Trust Framework for AI Agent Marketplaces (Accepted FSE 2026)
arXiv·medium signal
A three-layer decentralized reputation framework addressing how existing reputation mechanisms fail in agentic AI marketplaces because agents can game evaluations, competence doesn't transfer across contexts, and verification rigor varies. Separates task execution, reputation services, and tamper-proof persistence; introduces context-conditioned reputation cards preventing domain conflation; and provides a policy engine for adaptive verification escalation based on risk and uncertainty.