I have very mixed feelings about this one.
For utility steps that you write (run XSLT and save the secondary result elements to disk, for example), it would be convenient if you could put them all in $HOME/.mysteps.xpl (or something) and if the implementation would automatically load those steps every time. Then you could define ex:my-special-xslt and just use it.
Except you couldn't...
Because you'd need an explicit import of that library at the top of your pipeline. I mean, I suppose you could do that, but then if you always do that, you've made your pipelines less portable even if they don't use any of your steps. (And let's face it, the likelihood is that you're going to just import it as boilerplate in every new pipeline you write.)
I dunno.