By contrary, assume that p=ab+cd is a prime. In particular (b,c)=(b,d)=1, and ab≡−cd(modp). We get
0=b2(b2+bd−d2)−b2(a2+ac−c2)=b2(b2+bd−d2)−(ab)2−(ab)(bc)+b2c2≡(b2+c2)(b2+bd−d2)(modp)
Therefore, one of the numbers b2+c2 and b2+bd−d2 is divisible by p. If p∣b2+c2, then b2+c2=p since 0<b2+c2<2(ab+cd)=2p. This shows ab+cd=b2+c2 and then b∣c(c−d). But b,c are coprime, we get b∣c−d which is impossible.
If p∣b2+bd−d2, then b2+bd−d2=p since 0<b2+bd−d2<2b2<2ab<2p. This shows that ab+cd=b2+bd−d2 and then
b(a−b)=bd−cd−d2=d(b−c−d)
In particular, b−c−d>0 and b∣b−c−d (since b, d are coprime) which is impossible, again.
Hence, ab+cd must be a composite number.