How to read this: score is the gateway's risk score at submission; E[loss] is expected loss priced into the actuarial ledger; conduct codes (DRIFT, ESCALATION, VELOCITY, OFF-HOURS, DATA-EGRESS) feed the agent's A to D grade. Sample data mirrors the live sandbox scenarios. Nothing here is editable from this role, including by us: the chain would say so.