Runtime Enforcement & Resilience Layer

Deployment model: staged (shadow/evaluate -> controlled enforcement)
Type: capability layer (cross-product)
Frameworks: Threshold, Drift, Fidelity, Mandate, Meridian

Purpose

Runtime Enforcement & Resilience converts runtime scoring into staged, auditable policy actions while protecting against brittle automation.

It is applied across products, especially:

• Runtime Alignment Monitor
• Contract Assurance Monitor

Canonical Model

• Runtime Enforcement & Resilience = Threshold + Drift + Fidelity + Mandate + Meridian

Policy Modes

1. alert: detect and notify only.
2. shadow: simulate action outcomes without hard enforcement.
3. soft_block: controlled restrictions for scoped policies.
4. hard_block: strict enforcement for calibrated controls.

How It Works

1Ingest runtime score context

→

2Evaluate Threshold resilience confidence

→

3Apply policy mode rules

→

4Emit decision + rationale + audit payload

→

5Trigger human override path when required

Emits

policy actionpolicy_modedecision_reasonpolicy_versionaudit-ready event record

Integration Surfaces

• Consumes score outputs from:
  • POST /v1/runtime/analyze
  • POST /v1/contracts/obligations/score
  • GET /v1/score/{framework}/{entity_id}
• Uses signed webhooks for policy workflows:
  • score.updated
  • threshold.crossed
  • runtime.risk.elevated
  • enforcement.action.executed
  • enforcement.action.failed

Minimum Data Contract

• agent_id
• policy_mode (alert|shadow|soft_block|hard_block)
• policy profile references and policy version
• runtime score context (drift, fidelity, mandate, meridian, threshold)
• evidence lineage references for auditability

Rollout Guidance

1. Start with alert and shadow to establish precision/recall baselines.
2. Promote selected controls to soft_block with operational review.
3. Promote calibrated controls to scoped hard_block with governance approval.