The idea with voice control and virtual assistants is to be able to do certain things using a natural language, like if we were talking to another person. That's the power of it.
The problem with Custom Skills is that an "invocation name" is required to route the intent to a specific skill. This doesn't feel natural at all. For instance, if we had a real personal assistant, we would ask him/her to check our appointments, find a flight, etc and we would definitely not use phrases such as: "Check if I have any appointment for tomorrow from Fantastical" "Find me a flight from XXX to XXX for next Saturday using Sky Planner"
The "standard" intents to switch lights ON/OFF, set temperature, lock doors, etc. are an example of common sentences that are much more convenient to use now than before, when the user had to "manually" specify the skill to make it happen. Amazon is adding new "standardised" APIs for these common sentences, and that's better than nothing. However, this only happens as a reaction ("uh, many people are saying this, we probably want to make that easier"). As a Developer, I believe this is a turnoff: voice-controlled devices really need the freedom of natural language to explore new possibilities and experiment new features.
With all of that being said, I reckon that the problem here is to understand which Skill to use to handle a specific user request.
My proposal is to: 1. when the user says something that Alexa doesn't know how to process, the Skills are checked for one which supports the request 2a. If a Skill is matched, Alexa passes the intent to it 2b. If multiple Skills can handle the request, Alexa can ask which Skill to use among them and, the first time, ask if that Skill should always use that one for the future or keep asking. ... and that's is really. The current intent schemas and sample utterances could already contain all the needed info.
