Refer to the exhibit.

The correct choice is to use a distribute list outbound under the RIP process to stop EIGRP-learned prefixes from being advertised back into the RIP domain. In multipoint two-way redistribution, the major risk is route feedback: a route learned from one protocol is redistributed into another protocol and then reintroduced back into the original domain through another redistribution point. RIPv1 is especially limited because it is classful and has no native route tagging, so filtering must be carefully placed when route tags are unavailable. By filtering outbound advertisements under the RIP process, the design prevents routes that originated in the EIGRP domain from being sent into RIP where they could be relearned elsewhere and create loops or suboptimal paths. Inbound filtering on only one process does not reliably stop feedback at all redistribution points. A stronger modern design would use route tags with protocols that support them, but given the answer choices, outbound filtering toward RIP is the appropriate loop-prevention method. Reference topics: route redistribution, distribute lists, route feedback, RIPv1 limitations, EIGRP integration.