Offline receipt integrity vs. signer authority - where does the binding live?

[source-origin]

Read the security model — the part that keeps receipt integrity (integrity_valid) and signer trust (signer_trusted) as two separate verdicts, with the trusted key pinned out of band.

The seam I'd like your read on: when the receipt is verified by someone who was not the operator — a third party, later, on a different machine — what tells them which key should have signed a given decision? Integrity is checkable offline; trust needs a binding between the decision and the expected signer that does not itself come from the machine under audit.

Do you treat that binding as out-of-band only (a pinned public-key file), or does anything in the receipt / EvidencePack carry a verifiable statement of who was authorized to act — so a third party can check authority, not just signature integrity? The gap I keep hitting is the case where the recorder's own key is the only thing that says the recorder was allowed to decide.

(Adjacent context: we separate a receipt of an action from a verdict about it, and the binding between authority and receipt is the part we have not closed.)