Builds a deterministic sampling frame over historical AML alert dispositions for independent-validator review, plus a reviewer workload allocation. The statistical core (sample size, deterministic interval selection over your declared population hash) is not reimplemented here — it reuses the shipped attribute-sampling-plan kernel's calculation directly, the same systematic-sampling convention SOX 404 / ICFR control testing already uses. No randomness, so any reviewer can replay the exact same sample from the same declared inputs. This tool adds only the AML-specific layer: labeling the frame as a disposition sample and fanning the selected indices out round-robin across your reviewer roster.