A financial services firm needs a Claude-based agent to process portfolio rebalancing requests. Each request requires: (1) fetching current holdings, (2) fetching current market prices, (3) computing the rebalancing trades, and (4) submitting the trade orders. The team is designing the supervisor-worker topology. How should the work be partitioned?

D is correct because the supervisor-worker topology correctly maps task parallelism (holdings and price fetch are independent and can run concurrently) while maintaining sequential gating for dependent steps (computation requires both data fetches to complete). Having the supervisor validate intermediate results before advancing provides error containment — a failed price fetch should halt computation, not produce incorrect trades. Separating trade submission as a privileged worker adds an explicit authorization checkpoint. A is wrong because while a single agent can execute linear workflows, it cannot exploit the parallelism available in steps 1 and 2, and provides no supervision layer for error handling and validation. C is wrong because a naive fan-out where all four workers run simultaneously violates the dependency ordering — step 3 cannot run until steps 1 and 2 complete, and step 4 requires step 3's output. B is wrong because while human oversight of financial transactions is prudent, the architectural decision about AI agent involvement in step 4 is separate from the topology design question. The question asks about topology, not whether AI should submit trades.