Agent Rescue Desk — n8n Silent Workflow Watchdog Blueprint

{
  "type": "json",
  "example": {
    "ok": true,
    "limits": "This deterministic blueprint does not access, install, publish, or monitor an n8n instance.",
    "status": "complete",
    "privacy": "The submitted configuration was processed transiently. It was not stored or echoed.",
    "service": "Agent Rescue Desk n8n Silent Workflow Watchdog Blueprint",
    "version": "2026-07-27",
    "watcher": [
      {
        "id": "classify-overdue",
        "node": "Date & Time plus IF",
        "purpose": "Compare the last verified outcome with the guarded deadline.",
        "configuration": [
          "Classify the row overdue only when last_verified_at is more than 75 minutes old."
        ],
        "passCondition": "A current heartbeat routes healthy and a deliberately old heartbeat routes overdue.",
        "stopCondition": "Do not compare formatted local-time strings or mixed timezone bases."
      }
    ],
    "markdown": "# n8n silent-workflow watchdog blueprint",
    "nextStep": {
      "guideUrl": "https://agent-rescue-desk.netlify.app/n8n-workflow-stopped-silently-dead-man-switch",
      "rescueUrl": "https://marketplace.agent402.app/agents/6c54e490-72fc-47ba-bad1-3ac47b1b464a",
      "instruction": "Implement the producer and watcher in a non-production workflow, then run the controlled failure drill.",
      "rescuePriceUsd": 1000,
      "rescueServiceWindowHours": 12
    },
    "producer": [
      {
        "id": "verify-business-outcome",
        "node": "IF or Code",
        "purpose": "Prove the required downstream result instead of treating a green execution as success.",
        "configuration": [
          "Evaluate one deterministic, non-secret outcome contract after the final relevant side effect."
        ],
        "passCondition": "Exactly one independently observable outcome satisfies the documented success contract.",
        "stopCondition": "Do not advance the heartbeat when the business result is missing."
      }
    ],
    "alertPolicy": {
      "openingRule": "Open one incident when the first guarded deadline is missed.",
      "recoveryRule": "Close only after a newer verified-success heartbeat and emit exactly one recovery signal.",
      "duplicateRule": "Suppress additional opening alerts while incident_state remains open.",
      "escalationRule": "Escalate only on a separately documented longer threshold."
    },
    "failureDomain": {
      "action": "Verify the independent watcher cannot be disabled by the same host, process, database, or deployment.",
      "requirement": "At least one deadline expectation must run outside the watched n8n instance.",
      "sharedWithWatchedInstance": false
    },
    "stateContract": {
      "columns": [
        {
          "name": "last_verified_at",
          "type": "date",
          "purpose": "Time of the latest verified business outcome"
        }
      ],
      "uniqueKey": "watch_id",
      "storageModel": "One upserted current-state row per watched workflow plus bounded incident timestamps.",
      "retentionNote": "The current-state table remains approximately one row per watched workflow."
    },
    "normalizedContract": {
      "triggerType": "schedule",
      "graceMinutes": 15,
      "timezoneMode": "workflow",
      "triggerLabel": "Schedule Trigger",
      "cadenceMinutes": 60,
      "watcherLocation": "both",
      "expectedRunsPerDay": 24,
      "overdueAfterMinutes": 75,
      "watcherIntervalMinutes": 15,
      "successEvidenceSupplied": true
    },
    "verificationContract": {
      "completionRule": "Monitoring is ready only after the producer, watcher, deduplication, recovery, and independent failure-domain checks all pass.",
      "requiredEvidence": [
        "One controlled missed-run incident at the calculated threshold."
      ],
      "forbiddenShortcuts": [
        "Do not treat an in-instance watcher as proof that the same instance is available."
      ]
    },
    "controlledFailureDrill": [
      "Prevent one expected run in a safe test workflow.",
      "Verify exactly one incident opens and exactly one recovery follows."
    ]
  }
}