We use co-op students for (paid) four month terms. My number one concern when deciding to take an intern is "Am I going to get positive effort out of this exercise"? In other words, will the work the intern does significantly exceed the effort I have to expend training them?

So, the first thing I look for is evidence that they are quick on the uptake, self-motivated and so on. The second is familiarity with our basic computing environment - OS, editors, source control so that they can actually sit down and start doing real work as quickly as possible. You don't want to waste a couple of weeks having them learn to type 'ls' and use emacs.

Since we are a scientific organisation, I am not obsessed with specific language experience as long as the student has good experience with a "similar" language. This is because generally the students don't do "real" software engineering, they usually work in an existing framework, so they don't need advanced language skills.

So by having skills in a wide area, even if they are relatively shallow, you are already increasing your catchment area.

My gut reaction is "learn to use git", but I am not sure I could rationally defend it.