Hello Egor,
sounds like a great idea... and a workitem-exit would probably be even less transparent, as the logic would be found inside of the exit-implementation, instead of a container equation.
Probably, the completion-condition will only affect steps with a single outcome. User decision steps would need a "default" outcome, and I don't know, how that would actually work (without using a workitem-exit, that could probably do such a job, too).
So, for simple activities with one outcome, I like the completion-condition far more than what I had suggested myself :-)
Thumbs up.
Florin