---
title: "KYE™ Informed Approval™ — No blind approvals for AI-agent actions"
description: "KYE™ Approval Brief™ assembles actor, principal, authority, scope, state, and risk into a structured, signed brief — so human approvers decide informed"
url: https://kyeprotocol.com/solutions/informed-approval/
lang: en
source: "KYE Protocol"
---

> KYE™ Approval Brief™ assembles actor, principal, authority, scope, state, and risk into a structured, signed brief — so human approvers decide informed

KYE™ Approval Brief™

# No blind approvals for AI-agent actions.

When an AI agent's action needs human sign-off, the approver receives a structured, signed KYE™ Approval Brief™ — actor, principal, authority chain, scope, state, risk, recommended decision, recommended controls, consequence map, and the supporting scenario runs. Thirteen knowledge slots, every time. The decision is sealed into a KYE™ Approval Evidence Pack™ any auditor or regulator verifies offline.

1 · Why a brief

## Approvers were guessing. Now they know.

An approver presented with "Approve agent action #42" without context can only rubber-stamp or block. Both outcomes carry liability — and neither produces audit-grade evidence of what the approver knew at decision time. KYE™ Informed Approval™ ships a complete, structured brief: every fact the approver needs, captured in a snapshot, sealed alongside the decision.

- **Closes the audit gap.** The Approver Knowledge Snapshot records what was shown — facts, sources, and any questions asked.
- **Schema-conformant.** Brief, snapshot, decision, consequence map, and evidence pack all validate against their `v1` schemas.
- **Composer, not magic.** The V1 brief engine assembles the parts upstream callers (PEP-Express, gateway) supply. No hidden inference, no opinion — pure composition.

2 · The 13 required-knowledge items

## Every brief carries the same 13 slots.

Composition is deterministic — same inputs, same brief, same canonical bytes. The 13 slots cover the full who/what/why/where/when an approver needs to render a defensible decision.

### Actor

Acting party (agent / principal / service)

### Actor kind

agent · principal · service

### Principal

Entity on whose behalf the actor is acting

### Authority chain

Grant chain that authorises the action

### Scope

Bounded scope of the proposed action

### State

Tenant / principal / resource lifecycle states

### Proposed action

Schema, id, summary

### Risk level

From Scenario Testing™

### Recommended decision

allow / allow\_with\_constraints / require\_approval / deny

### Recommended controls

Controls implied by stressor evaluation

### Consequence map

Intended + possible-unintended consequences

### Knowledge snapshot

Facts shown to the approver, captured

### Validity window

not\_before / not\_after for the brief

3 · The pipeline

## From action request to sealed evidence.

### Action Request

— risk\_level + recommended\_decision Scenario / Stress Check

### Consequence Map

— intended + possible-unintended ( ) kye.consequence\_map.v1

### Approval Brief

— composer assembles 13 slots ( ) kye.approval\_brief.v1

### Knowledge Snapshot

— facts shown to approver ( ) kye.approver\_knowledge\_snapshot.v1

### Approval Decision

— approver decides ( ) kye.approval\_decision.v1

### Evidence Pack™

— sealed + ed25519-signed ( ) kye.approval\_evidence\_pack.v1

### PDP

— decision flows back to enforcement

**Why this order.** Scenario before consequence; consequence before brief; brief before snapshot; snapshot before decision; decision before seal. Each stage feeds the next via canonical references — no stage receives information the prior stage did not authorise.

4 · Sample Approval Brief

## A real, schema-valid approval brief.

Below: a real `kye.approval_brief.v1` example for a payment-agent scope-expansion approval. Click to expand. The brief carries references to the knowledge snapshot, the consequence map, and the supporting scenario run — all sealed into the evidence pack at decision time.

Sample `kye.approval_brief.v1`

```
{
  "schema": "kye.approval_brief.v1",
  "id": "kye:approval-brief:payment-agent-scope-expansion.2026-05-14",
  "version": 1,
  "proposed_action": {
    "action_id": "kye:proposed-action:expand-payment-agent-limit-50k",
    "action_schema": "kye.proposed_action.v1",
    "summary": "Expand acme-bank payment orchestrator per-transaction limit from $10k to $50k on FedNow rail, effective 2026-06-01."
  },
  "approver": "kye:person:acme-bank.cro",
  "knowledge_snapshot_id": "kye:approver-knowledge-snapshot:payment-agent-scope-expansion.cro.2026-05-14",
  "consequence_map_id":    "kye:consequence-map:payment-agent-scope-expansion.2026-05-14",
  "supporting_scenario_runs": [
    "kye:scenario-run:payment-agent-scope-expansion.2026-05-14"
  ],
  "recommended_decision": "approved_with_restrictions",
  "validity": {
    "not_before": "2026-05-14T10:00:00Z",
    "not_after":  "2026-05-21T10:00:00Z"
  },
  "issued_at": "2026-05-14T09:30:00Z",
  "issued_by": "kye:agent:approval-brief-assembler",
  "signature": {
    "alg": "EdDSA",
    "kid": "kye:key:approval-brief-2026-05",
    "value_b64": "TG9yZW1hcHByb3ZhbEJyaWVmU2lnbmF0dXJlRXhhbXBsZV8x"
  }
}
```

5 · Next

## Run the scenario, then issue the brief.

Scenario Testing™ emits the risk evidence the brief composes from. Together they close the loop between simulation and sign-off.

[Scenario Testing™](https://kyeprotocol.com/solutions/scenario-testing/) [Evidence Pack™](https://kyeprotocol.com/evidence-pack/) [Assurance Card™](https://kyeprotocol.com/assurance-card/)

Canonical KYE™ surfaces referenced on this page: [KYE Protocol™](https://kyeprotocol.com/).