0% read on this device
Browse the curriculum

Start Here

Vision

First Principles

Operating Model

Domain Model

Agent Factory

Runtime Architecture

AI Engineering

Autonomous Workflows

Verification & Delivery

Factory Platform

Quality Engineering

Security & Governance

Case Studies

Labs

Interview Practice

Research Journal

Reference

Curriculum/Security & Governance/A focused view of boundaries, contracts, state, authority, failure paths, and tradeoffs drawn from this chapter.
Security & Governance5 min readchapterQuick Read

Authority, Autonomy, and Emergency Control

Define how authority is delegated, exercised, reduced, revoked, contained, and restored.

Status: Review readyRisk: criticalLifecycle: intent · execute · verify · deliver · learnContent reviewed 2026-08-30Maturity guide →
Claim boundaryThis is curriculum guidance. It does not by itself prove a production implementation.
architecture mode

A focused view of boundaries, contracts, state, authority, failure paths, and tradeoffs drawn from this chapter.

4. Command and event contracts

Every emergency command includes command ID, subject selector, requested action, actor identity, delegated role, reason, severity, expected version, idempotency key, issued time, deadline, and notification policy. The control plane records acceptance or rejection before dispatch. Components emit control.requested, control.acknowledged, control.enforced, control.failed, and control.verified events with one correlation key.

Acknowledgement is not enforcement. Enforcement is not verified safety. Verification checks scheduling, running work, credentials, queues, network access, side effects, artifacts, and downstream systems.

5. Control sequence

  1. Detect or receive a credible report.
  2. Resolve subject and blast radius from inventory and live state.
  3. Authorize the narrowest sufficient emergency action.
  4. Record the command durably before dispatch where possible.
  5. Fan out through independent control paths to scheduler, orchestrator, tool gateway, credential broker, and deployment system.
  6. Preserve attempts, events, artifacts, and relevant volatile state.
  7. Reconcile acknowledgement with observed enforcement.
  8. Escalate on timeout; use a broader control only when justified.
  9. Diagnose and choose compensate, rollback, failover, retire, or repair.
  10. Independently verify recovery before restoring authority.

7. Failure and recovery matrix

FailureRequired behaviorEvidence
Orchestrator unavailableIndependent scheduler/gateway control path denies new effectsGateway denials and queue freeze
Worker ignores pauseLease expires; gateway denies tool calls; worker isolatedLease, identity, and network events
Duplicate cancel commandIdempotent terminal result; no duplicate compensationCommand deduplication record
Revocation cache staleCritical tools fail closed after bounded cache TTLCache version and denial test
Partial external side effectRecord uncertainty, stop retry, reconcile with providerProvider idempotency key and reconciliation result
Control operator unavailableDelegated backup assumes authority under policyOn-call transfer and signed decision
Recovery uses changed contextResume denied; create new manifest or explicit replanContext/version comparison

9. Human override and dual control

Every high-impact action has a named override point and accountable owner. Dual control applies when one compromised or mistaken identity could create an irreversible or privileged effect. Emergency containment may use a single delegated incident authority when delay is more dangerous, but restoration of broad authority requires the defined approval and independent verification.

External review

Review this chapter.

Challenge a claim, boundary, missing failure mode, unclear term, or unsupported evidence statement.

  • Claim
  • Boundary
  • Failure
  • Evidence