Yeah. I like it for how well it seems to model most REST APIs. But as you can see, it's not as simple as inputs and outputs. And that's why attaching something so general to model any REST API (not even as general as any web API) as a prerequisite to web pipes would probably really hurt adoption.
That said, if something like RestDoc was widespread, we'd maybe be able to skip block definitions. We'd still end up having to impose some constraints and people would likely still have to be aware of being "block compatible". For example, the fact that pipeline executors treat multiple output objects a particular way ... that gives semantic meaning to the response that's specific to webpipes, and might not work for just any API.
Wasn't Roy Fielding big on self-describing resources? Has he been involved in anything like RestDoc? I think Twilio's self-describing API was based on whatever that was...