At-risk customer alert

{
  "jimothyPipeline": 1,
  "name": "At-risk customer alert",
  "description": "Scores tickets and CSAT ratings for churn risk and alerts the account owner when it’s high.",
  "icon": "workflow",
  "color": "#ca8a04",
  "repo": {
    "mode": "scratch"
  },
  "concurrency": 5,
  "variables": {
    "savePlays": "Call from a VP within 24 hours. Engineering ETA on the underlying bug. One month credit (up to $5,000 ARR accounts). Dedicated onboarding session. Extended trial of a higher plan.",
    "slackWebhookUrl": ""
  },
  "steps": [
    {
      "id": "score",
      "name": "Score churn risk",
      "type": "agent",
      "description": "Rates the risk as high, medium or low with the signals behind it.",
      "harnessId": "anthropic-api",
      "model": "claude-haiku-4-5",
      "timeoutMinutes": 3,
      "retries": 1,
      "passPattern": "^\\s*RISK:\\s*(high|medium|low)\\b",
      "prompt": "Customer: {{issue.raw.customer | default:\"(unknown)\"}}, plan {{issue.raw.plan | default:\"unknown\"}}, ARR {{issue.raw.arr | default:\"unknown\"}}\nRecent history: {{issue.raw.history | default:\"(none sent)\"}}\n\nNew message ({{issue.key}}): {{issue.title}}\n{{issue.description | default:\"(empty)\"}}\n\nRate the risk that this customer churns. High means explicit talk of leaving, cancelling, evaluating alternatives or escalating, or repeated failures with strong frustration. Answer in exactly this format:\nRISK: high, medium or low\nSIGNALS: the phrases or facts that drove the rating"
    },
    {
      "id": "high",
      "name": "High risk?",
      "type": "condition",
      "description": "Stops here unless the risk is high.",
      "dependsOn": [
        "score"
      ],
      "condition": {
        "match": "all",
        "rules": [
          {
            "value": "{{steps.score.output}}",
            "op": "matches",
            "compare": "RISK:\\s*high"
          }
        ]
      }
    },
    {
      "id": "brief",
      "name": "Brief the owner",
      "type": "agent",
      "description": "Writes the alert: what’s wrong, what the account is worth, a save play and a draft reply.",
      "harnessId": "anthropic-api",
      "model": "claude-sonnet-5",
      "dependsOn": [
        "high"
      ],
      "timeoutMinutes": 5,
      "prompt": "{{steps.score.output}}\n\nCustomer: {{issue.raw.customer | default:\"(unknown)\"}}, plan {{issue.raw.plan | default:\"unknown\"}}, ARR {{issue.raw.arr | default:\"unknown\"}}, owner {{issue.raw.accountOwner | default:\"unassigned\"}}\nHistory: {{issue.raw.history | default:\"(none sent)\"}}\nMessage ({{issue.key}}): {{issue.title}}\n{{issue.description}}\n\nSave plays we can offer: {{vars.savePlays}}\n\nWrite a Slack alert (*bold*, - bullets), under 200 words: *At risk: <customer> (<ARR>)*, owner, what's wrong in two lines, the save play you recommend from the list and why, and a short draft reply for the owner to send personally."
    },
    {
      "id": "post",
      "name": "Alert in Slack",
      "type": "shell",
      "description": "Posts the alert for the account owner.",
      "dependsOn": [
        "brief"
      ],
      "timeoutMinutes": 2,
      "command": "printf '{\"text\":%s}' {{steps.brief.output | json}} | curl -fsS -X POST -H 'Content-Type: application/json' --data @- {{vars.slackWebhookUrl}}"
    }
  ]
}