For the $k=4$ case, one can simply note that
$$n^4 + 4 = (n^2 - 2n + 2)(n^2 + 2n + 2)$$
The motivation is realizing that you have something that looks like it could factor as two factors of $n^2 + ?n + 2$, since $(n^2)^2 = n^4$ and $2^2 = 4$, and trying to determine what that unknown quantity is. One observes that, letting that quantity be variables $b,c$ in two factors, that
$$(n^2 + bn + 2)(n^2 + cn + 2) = n^4 + n^3 (b+c) + n^2 (bc + 4) + n(2b+2c) + 4$$
So one requires that
$$\begin{cases}
b+c = 0 \\
bc + 4 = 0 \\
2b + 2c = 0
\end{cases}$$
though the latter equation is redundant. So you end up having two numbers whose sum is zero, and whose product is $-4$. Some trial and error readily let's one see that $b = -2$ and $c=2$. (Or I suppose you could try substitution or another method.)
That said, I'm not really sure about generalizations on the $n^k + k$ case.
- Obviously $k=1$ is a dud. Take $n=6$. ($7$ is prime.)
- In the case $k=2$, take $n=3$. ($11$ is prime.)
- In the case $k=3$, take $n=2$. ($11$ is prime.)
- In the case $k=5$, take $n=2$. ($37$ is prime.)
- In the case $k=6$, not sure. (Nothing sticks out as one to me with small cases.)
- In the case $k=7$, take $n=4$. ($16,391$ is prime.)
- In the case $k=8$, take $n=3$. ($6,569$ is prime.)
- In the case $k=9$, take $n=2$. ($521$ is prime.)
- In the case $k=10$, not sure. (Nothing sticks out as one to me with small cases.)
So I know it's true for $k=4$, and maybe $6$ and $10$, but I'm not sure.
Not really up for trying more cases, but I'm not sure of a pattern offhand.