assets/runtime-health-manifest.yaml

schema_version: frontier.machine.health.v1
deployment_id: replace-deployment
checked_at: "2026-08-06T00:00:00Z"
layers:
  process:
    purpose: "The service process is alive and exposing a heartbeat."
    checks:
      - id: process-heartbeat
        status: pass # pass | fail | unknown
        critical: true
        observed_at: "2026-08-06T00:00:00Z"
        stale_after_seconds: 60
        summary: "Process heartbeat responded."
  scheduler:
    purpose: "The scheduler can claim due work and advance queues."
    checks:
      - id: scheduler-claim
        status: pass
        critical: true
        observed_at: "2026-08-06T00:00:00Z"
        stale_after_seconds: 120
        summary: "Scheduler claimed and acknowledged due work."
  execution:
    purpose: "Workers can perform representative real work, including provider auth and credit paths."
    checks:
      - id: provider-smoke
        status: pass
        critical: true
        observed_at: "2026-08-06T00:00:00Z"
        stale_after_seconds: 300
        summary: "Representative worker completed a provider-backed smoke task."
  governance:
    purpose: "The gate can deny, allow, halt, and record decisions independently."
    checks:
      - id: gate-deny
        status: pass
        critical: true
        observed_at: "2026-08-06T00:00:00Z"
        stale_after_seconds: 300
        summary: "Governance gate denied an unauthorized mutation."
aggregate_policy:
  status: fail_closed
  rule: "can_mutate is true only when process, scheduler, execution, and governance all pass fresh critical checks."
  warning: "A live process heartbeat must not mask a dead scheduler, dead worker pool, provider auth or credit failure, missing verifier, unratified contract, active override, no-ACK halt, or failed governance gate."