I suspect (because I've had not double checked) what @paul driver is correct. However, it may be unclear to you what and why Paul suggested what he did.
Routing looks at an incoming packet and compares that packet's destination address with all its known networks (i.e. their prefixes) and aggregates of such, or, and lastly, a "none of the above" (the latter is not done, by default). Matching occurs from the most specific match to the least specific match.
In your diagram you have five networks containing hosts and five networks as transits between routers.
For packets to flow between devices, each router needs to know where to direct a received packet. (If it doesn't, the packet is dropped.)
By default, routers know all the networks directly connected to them, but need to (somehow) be informed of networks not directly connected so they can forward the packet to the correct egress interface and to, sometimes, a known IP to forward packets to.
So, Paul "informs" the central router of all five (not directly connected) host networks and what IP to forward those packets to (the latter implying the router knows how to get to those IPs, which it does, as they are all directly connected networks.) (BTW, Paul is assuming all those host networks are /24, but they could be smaller or larger. In fact, using the old Classful network rules, I believe two are Class As, /8, one a Class B, /16, and two Class Cs, /24).
For the routers, either they need to be "informed" of the four other host networks, going via the central router, or, what Paul has done, is "inform" them, send "unknown" destinations to the central router. I.e. one "wildcard" route on each host edge router replaces four routes that you would otherwise need. (BTW, this would be the usual approach for a hub/spoke network.)
If you're still confused, post additional questions, because, the real goal, for such an assigned project, is to "learn" routing.
