Techniques for exposing virtual assistant skills as authorable content items interpretable across multiple platforms are disclosed. In an embodiment, a query system comprises an execution engine and a plurality of wrapper services including (a) a first wrapper service for converting messages between a first protocol of a first virtual assistant platform and a target protocol and (b) a second wrapper service for converting messages between a second protocol of a second virtual assistant platform and the target protocol. The query system may receive a specification for a skill and requests from the plurality of virtual assistant platforms that invoke the skill. The wrapper services convert the requests to a target protocol consumable by the execution engine and convert responses from the execution engine a protocol consumable by the virtual assistant platform through which the request originated.