Is a separate JSON action format actually useful?

I’ve been experimenting with a small JSON action format called QUORACT.

It uses five fields: object, intent, authority, lifetime, and result.

The idea is to keep the description of an action separate from how it’s transported or executed.

I have a basic specification and 12 public test cases so far. Authorization and replay protection aren’t implemented by the public profile.

I’m still trying to figure out whether this is useful enough to justify another format, especially with MCP and A2A already around.

Has anyone dealt with a similar problem? Would you keep this as a separate layer or just use existing protocols?

Code and spec: https://github.com/silentxtrader/QUORACT