Architecture Patterns¶
These pattern stubs are starting points for future expansion. Each pattern should eventually include intent, context, forces, implementation notes, risks, and example signals.
Drift Detector¶
Identifies movement away from expected behaviour, policy, state, cost, quality, or outcome. A drift detector should distinguish normal adaptation from movement that weakens coherence.
Governance Checkpoint¶
Evaluates policy, authority, risk, context, or constraints before execution continues. A checkpoint can approve, reject, hold, or escalate.
Runtime Veto¶
Stops an action while the system is running. A runtime veto should record the triggering condition, relevant context, authority, and next available path.
Coherence Boundary¶
Defines where coherence expectations apply and who is responsible for maintaining them. A boundary may surround an agent group, workflow, domain, organisation, or federation.
Policy-Aware Delegation¶
Carries policy, authority, constraints, and context when work is delegated. The receiving actor should know not only what to do, but also what bounds apply.
Federation Health Signal¶
Reports whether a federated set of agents or systems remains within acceptable operating bounds. Signals may include conflict rate, stale context, veto rate, escalation load, cost variance, and unresolved drift.
Escalation Router¶
Routes ambiguous, risky, conflicting, or unauthorised situations to the right actor or control plane. Routing should be explicit, auditable, and connected to authority.
Audit Trail Spine¶
Provides a consistent record across agents, tools, workflows, and control-plane decisions. The spine should connect intent, context, authority, action, policy checks, state changes, and outcome.