Maybe app actions are only available on Pixel devices.
I want every app to start thinking about voice ux, but that's probably down to the individual apps, and you're right that some do. I can play some things on Stitcher, say, though I can't remember if I have to use podcast names specifically or if I can use my playlist names.
Some Google apps don't seem to support it though.
Ok Google, log my weight on Google fit as X? Search results page.
Ok Google, play (my playlist name) on Google Play: plays unrelated radio station inspired by my playlist name, with sometimes comical results.
I'm sure it will improve with time.
I think once somebody finds a good API for that, it will be a watershed moment.
Amazon Alexa has apps called "skills", but they expose their own voice interface which is always inconsistent.
But I don't know if there's enough incentive for apps to add an assistant interface. How easy is it to add commands on your app that it will accept via assistant? Are there common APIs or structures for this to shift the incentives, allow it to proliferate app by app, rather than waiting on the assistant developers to manually do each one?
After that we need to go back to Aza Raskin's project, uh, Ubiquity? Anyone could build little scripts, verbs, that helped build a natural language input from the ground up. That's how we get to stuff like, "Hey Google, take a screenshot of news.ycombinator.com and email it to Joe."