njguliyev and Hagen von Eitzen have shown the correct approach. I just want to belabor a point, in case it was too subtle. :) By definition, a series converges if its sequence of partial sums converges. That is THE definition of series convergence.
All of the tests you've studied -- the term test, the ratio test, the root test, the condensation test, the alternating series test, even the Cauchy criterion -- are consequences of this definition. So if an exam says "Prove directly from the definition of series convergence", it is asking you NOT to apply any of the tests!
The purpose of the question is not to see whether you can solve the mathematical problem to determine whether a particular series converges. The purpose is to see whether you know the definition and whether you can apply it, without calling upon the most powerful tools you've been given. If you use any "test", you won't get full credit.
(As an aside, the ratio test happens to be inconclusive for this series anyway!)
Unfortunately, there's still a bit of a ambiguity. The sequence of partial sums is $(-1,0,-1,0,\ldots)$. Clearly it doesn't converge, but are you required to prove that as well? If so, are you required to argue directly from the epsilon-delta definition of a limit? Or are you allowed to call upon the theorems about sequences that you've studied? For example, is it good enough to observe that the sequence has two subsequences that converge to different limits? Only the grader for the exam knows for sure.
During a real exam, you could write down "This sequence doesn't converge" and then move on to the next question. When you finish the exam, if you still have time remaining, you can go back and provide a proof, just to be safe.