Count I — Technically successful
The action completed without error and returned the expected result. Admitted. Success is the floor, not the defence — a clean execution says nothing about the right to execute. not in dispute
KYE Protocol™
Authority Court: the agent on trial.
Picture a courtroom. In the dock is not the agent, but a single thing it did — one consequential action it already took. The action was fast, it was fluent, and it worked. That is not the question on trial. The charge is Authority Finality™: was the institution's mandate for this action valid, in scope, and settled at the moment it acted? A successful action is not a legitimate one, and this is the court that tells them apart — turning an argument that used to take days into a verdict you can read in minutes.
This is a dramatization. There is no real tribunal, no legal proceeding, and no verdict in law — the courtroom is a metaphor for one engineering question KYE Protocol™ answers at the action boundary: was the actor authorised?
The docket
The defendant is the action, not the agent.
You are not here to decide whether the agent is clever, well-built, or well-behaved in general — you are here to judge one act it took against one principal, for one purpose, at one moment.
An agent that answers a thousand questions correctly and then moves a payment it was never empowered to move has committed exactly one act worth trying. So the court narrows to that single action: a payment released, a policy bound, a record disclosed, a case escalated. Every other action gets its own hearing. This is the discipline KYE Protocol™ enforces — authority is judged per action, at the boundary, never averaged across a run.
The charge sheet
Four counts the court reads aloud.
The prosecution does not dispute that the action ran. It asks 4 questions in order, because an action can pass the first three tests every engineer usually checks and still fail the one at t=0 that matters.
Count II — Policy compliant
The action satisfied the written policy the team compiled into code. Useful, but a rule being satisfiable is not the same as anyone holding the authority to act on it — that is Action Admissibility™, one stage, not the whole verdict. necessary, not sufficient
Count III — Authority valid
Did this principal actually hold delegated authority for this purpose, within scope and limits, from a source empowered to grant it? This is where most actions that "succeeded" are found wanting — and where KYE Protocol™ returns an admit-or-deny verdict. the real charge
Count IV — Effect final
Was the authority still live at t=0 — the instant before the world changed — and sealed so it can be replayed later? Authority that lapsed a second earlier makes the effect indefensible, however good the intent. settled at t=0
The defence that fails
Four arguments the court will not accept.
The defence offers the four exhibits teams most often mistake for authority — an annual SOC 2 attestation, an access-control allow, a full trace, a policy-owner sign-off. Each is real and useful; none of them, on its own, ever proves the action was authorised.
| The exhibit offered | Why the court rejects it |
|---|---|
| "The policy owner approved it." | A policy owner is not the source of the mandate. Owning the rule is not the same as being empowered to grant the authority the action relied on. |
| "We passed our periodic attestation." | A quarterly or annual attestation certifies a control at a point in time. It is not a per-action approval for the specific act on trial. |
| "Our RBAC and ABAC allowed it." | Access control decides whether the caller may reach the system. It is not the external mandate that says this principal may take this action, here, now. |
| "We have the full trace." | A trace proves what happened. It is silent on legitimacy — it records the act without ever recording the grant behind it. |
Each argument answers a real question — reachability, control coverage, observability — but none answers the charge. The court keeps returning to Count III and Count IV, because authority and finality are the two things none of these exhibits contain.
The evidence
The record that can be replayed, not just recited.
A verdict is only as good as the evidence behind it, so KYE Protocol™ seals the basis of every decision into a record a stranger can re-check without trusting the court.
For the action on trial, the decision is packed into an Evidence Pack™ — the principal, the delegated authority, the purpose, the scope, the limits, and the decision itself, captured with a Decision Map™ of the reasons. It is signed as a Replay-Proof™ that a regulator or counterparty can verify from public keys alone, offline, long after the hearing. That is the difference between a court that asserts a verdict and one that hands you the file to check for yourself — and it is why exam-prep that once took days can be produced in minutes.
The verdict
Authorised, or merely successful.
The court returns to the only question that ever mattered, and it has a name.
Verdict · Authority Court
An action is acquitted only when its authority was valid, in scope, live at t=0, and sealed as evidence. Everything else — fast, fluent, error-free, policy-passing — is merely successful. The name of the acquittal is Authority Finality™.
Honest scope
A metaphor, and one honest claim.
KYE Protocol™ proves authority, not truth — and this courtroom is a way to picture that, not a legal proceeding.
The trial does not decide whether the agent's answer was correct, its model well-calibrated, or its output wise; that is the work of evaluation, and it stays necessary. KYE Protocol™ proves exactly one thing, and proves it completely: whether the actor was authorised to take the action. The dramatization is a lens on that boundary — no tribunal, no ruling in law, just the single engineering question a dashboard never answers, made concrete.
Canonical KYE™ surfaces referenced on this page: KYE Protocol™.