Steam (gas-phase H2O) is very commonly used as the working fluid for external-combustion heat engines. It seems to me that it ought to be a poor choice, because:
Unless I am seriously confused about something, you have no choice but to throw the working fluid's heat of vaporization away every time around a Rankine cycle. (You might be able to recover some of it with an economizer, but that's a bolt-on.) This drawback seems like it ought to apply to all heat engine designs involving a gas-liquid phase transition, but water should be particularly bad in this context because its heat of vaporization is so large.
Relatedly, phase transitions are a source of inherent irreversibility and thus efficiency loss, aren't they?
Water also has a very high specific heat capacity, both as liquid and as gas; I'm not sure about this part, but my best guess is that you want a working fluid with a low specific heat capacity, so that as much of its thermal energy as possible will be translation-kinetic energy of the molecules, and therefore directly convertible to motion.
Water condenses to liquid at 100°C, which is well above the temperatures of readily-available cold sinks (ambient air usually from 10–30°C; cold water is often available at 5°C or so); shouldn't a working fluid that could be expanded all the way down to the cold-sink temperature allow greater efficiencies?
"Wet" steam is reactive enough to cause a variety of engineering headaches, e.g. turbine blade erosion.
Instead, the abstractly correct choice would seem to be a permanent, inert gas; dry nitrogen and dry air are readily available for cheap and should be plenty inert enough for most applications (maybe at very high temperatures you go to helium or argon). But "hot air" reciprocating engines are rare, and nobody tries to build gas turbines that expand the gas all the way down to ambient—instead they use the exhaust as the hot reservoir for a steam engine!
There must be something I'm missing. Please explain.