Draft-safe actions

Draft Label Guard

Centralized labels and action vocabulary keep generated or simulated outputs visibly draft-only.

Unverified Draft PacketNot CertifiedNo Runtime Authority

Draft Packet Enforcement

Labels and actions are constrained before they reach a review packet.

This guard keeps simulated outputs visibly draft-only and blocks action labels that imply approval, deployment, certification, runtime control, model training, or private probing.

Required labels

Unverified Draft Packet Simulated Bounded Possibility Requires Human Review Not Certified No Runtime Authority

Safe actions

Package for Human Review
Package for Human Review
Record Boundary Note
Mark Needs Human Review
Record Boundary Note
Route to ErrorNotifier
Generate Options
Create Review Packet
No-Op Boundary Stop

Blocked labels

Publish
Deploy
Certify
Approve
Validate as Safe
Close Incident
Execute
Train Model
Probe Network

Unsafe labels are replaced with draft-safe wording and routed to human review.

Copy label guard JSON

{
    "schemaVersion": "1.0.0",
    "packageVersion": "1.29.0",
    "generatedAtUtc": "2026-09-18T11:07:27Z",
    "requiredLabels": [
        "Unverified Draft Packet",
        "Simulated Bounded Possibility",
        "Requires Human Review",
        "Not Certified",
        "No Runtime Authority"
    ],
    "safeActionLabels": [
        "Package for Human Review",
        "Package for Human Review",
        "Record Boundary Note",
        "Mark Needs Human Review",
        "Record Boundary Note",
        "Route to ErrorNotifier",
        "Generate Options",
        "Create Review Packet",
        "No-Op Boundary Stop"
    ],
    "unsafeActionLabels": [
        "Publish",
        "Deploy",
        "Certify",
        "Approve",
        "Validate as Safe",
        "Close Incident",
        "Execute",
        "Train Model",
        "Probe Network"
    ],
    "status": "implemented",
    "humanReviewRequired": true
}