Backbone for Assistants like Alexa, Siri, OK Google and Co.

Skills, Workflows or what ever they are called

Alexa uses specific skills to hold the logic for voice commands. Siri an OK Google do the same. Whenever a intent is identified, the right logic must be executed. The problem is, some developer must thought about the logic- Because, without a skill, no action is fullfilled by the assistant.

How to change it?

Why not having a backbone system that generates skills on the fly, in order to fullfill the intent the user has? Maybe through building blocks for the skills, through history and comparison of existing skills or by asking the user about specific steps in the overall task.

What do you guys think. How can assistants escape the trap, that every action must be "hard coded" into a skill?