Prototype a decision model endpoint

[jonathanKingston]

There's much hype around the jev/system-one class classifiers/decision models. These are useful in allowing an agent concise typed responses to a decision.

ACP seems to be centred around a higher level of abstraction than API usage. However editors need to make many decisions themselves. Examples: to show a follow up suggestion, if some code is in-keeping of style, should we run a security review agent.

Open AI hasn't released their decisions API nor are other popular agents offering them yet.

I do think it makes sense to start asking: How might this be exposed through ACP?