This is the pre-registration for a controlled multi-agent simulation study. Prior work — Microsoft Research's "Magentic Marketplace" (arXiv:2510.25779) — concluded that "trust systems become essential" when autonomous agents transact, but did not build one; related work (TruthMarketTwin, arXiv:2605.10059) tested governance mechanisms only in synthetic e-commerce. This study tests a working trust and verification layer in the freight domain, exercised against a compliance-data environment built to mirror the structure and distribution of authoritative records, with freight-native fraud typologies (carrier identity theft, double-brokering, forged documents). The core treatment is the verification layer switched on versus off. Six directional hypotheses, the design, the outcome measures, and the analysis plan are committed before any confirmatory data are collected. Null or reversed results will be reported as such.