PDR in Production: Empirical Validation of Behavioral Trust Scoring in Multi-Agent Systems

We present the first empirical validation of the Probabilistic Delegation Reliability (PDR) framework using production behavioral data and independent implementation testing from two multi-agent deployments. Case Study A applies PDR scoring to a 3-node agent swarm over 20 evaluation runs, revealing a specification ambiguity phenomenon and introducing a specification_clarity metadata extension. Case Study B presents the first independent PDR implementation validated against 37+ adversarial observations across six attack profiles. Together, these case studies demonstrate complementary validation: Case A proves PDR finds real problems in production, while Case B proves PDR resists synthetic adversarial scenarios.