Let $k$ be a field. How to prove that every finitely generated $k[x]/(x^2)$-module is a direct sum of copies of $k[x]/(x^2)$ and $k[x]/(x)$?
This follows from a more general result by Nakayama on modules over uniserial Artinian rings, but I wonder if there is a direct proof which does not use much theory. Actually the theorem applies to all modules, but I am only interested in finitely generated modules here. The tag "representation theory" comes from the interpretation $k[x]/(x^2) \cong k[\Sigma_2]$ when $\mathrm{char}(k)=2$. Thus, the goal is to classify all representations of $\Sigma_2$ in characteristic $2$.
Edit: Sorry, this was a silly question. Since $k[x]$ is a PID, finitely generated $k[x]$-modules have a well-known classification. One just has to look at those killed by $x^2$, and we are done.
I am not sure what you are allowed to use. This is a simple consequence of the fact that $A=k[x]/(x^2)$ is an injective $A$-module. If $M$ is any finitely generated $A$-module and if it has an element $m$ with $\mathrm{Ann}\, m=0$, then we get an inclusion $A\to M$, which splits. Thus we are reduced to the case when all elements in $M$ have non-zero annihilator. But, then $xM=0$ and the rest is clear.