{
  "jimothyPipeline": 1,
  "name": "Decision red team",
  "description": "Frames a decision, has three reviewers attack it in parallel, and writes a decision memo with a recommendation.",
  "icon": "workflow",
  "color": "#8b5cf6",
  "repo": {
    "mode": "scratch"
  },
  "concurrency": 2,
  "variables": {
    "company": "Acme, a 120-person B2B software company with $18M ARR",
    "strategy": "Reach profitability by Q4 while growing net revenue retention. Mid-market finance teams are the core segment; enterprise is opportunistic."
  },
  "steps": [
    {
      "id": "frame",
      "name": "Frame the decision",
      "type": "agent",
      "description": "Restates the decision, the options and every assumption it depends on.",
      "harnessId": "anthropic-api",
      "model": "claude-sonnet-5",
      "timeoutMinutes": 5,
      "prompt": "Company: {{vars.company}}\nStrategy: {{vars.strategy}}\n\nProposal: {{issue.title}}\n{{issue.description}}\n\nFrame this decision for reviewers, in Markdown:\n## Decision (one sentence, as a yes/no question)\n## Options (including doing nothing)\n## Assumptions (numbered; every claim the proposal depends on, stated so it could be proven false)\n## Known numbers (only what the proposal gives)\n## Missing information"
    },
    {
      "id": "finance",
      "name": "Financial review",
      "type": "agent",
      "description": "Attacks the costs, payback and cash impact, and finds the assumptions the case is most sensitive to.",
      "harnessId": "anthropic-api",
      "model": "claude-opus-5",
      "dependsOn": [
        "frame"
      ],
      "timeoutMinutes": 10,
      "systemPrompt": "You are a skeptical CFO. Strategy: {{vars.strategy}}",
      "prompt": "{{steps.frame.output}}\n\nAttack this from the financial side: total cost including the costs the proposal leaves out, payback period, cash and runway impact, the opportunity cost against the strategy, and which numbered assumptions the case is most sensitive to. Show any arithmetic. Under 400 words."
    },
    {
      "id": "customer",
      "name": "Customer review",
      "type": "agent",
      "description": "Attacks demand: would customers care, what competitors would do, what it signals.",
      "harnessId": "anthropic-api",
      "model": "claude-sonnet-5",
      "dependsOn": [
        "frame"
      ],
      "timeoutMinutes": 10,
      "systemPrompt": "You represent the company’s customers and the market. Strategy: {{vars.strategy}}",
      "prompt": "{{steps.frame.output}}\n\nAttack this from the customer and market side: would customers notice or care, what competitors would do, what it signals, and which numbered assumptions about demand are weakest. Under 400 words."
    },
    {
      "id": "risk",
      "name": "Execution review",
      "type": "agent",
      "description": "Runs a pre-mortem and proposes the cheapest 30-day test of the riskiest assumption.",
      "harnessId": "anthropic-api",
      "model": "claude-opus-5",
      "dependsOn": [
        "frame"
      ],
      "timeoutMinutes": 10,
      "systemPrompt": "You are the operator who would have to make this work. Strategy: {{vars.strategy}}",
      "prompt": "{{steps.frame.output}}\n\nRun a pre-mortem: it's a year from now and this failed. Give the three most likely reasons, what the team would have to stop doing to make room for it, the hardest dependency, and the cheapest experiment that would test the riskiest numbered assumption within 30 days. Under 400 words."
    },
    {
      "id": "memo",
      "name": "Decision memo",
      "type": "agent",
      "description": "Weighs the three reviews and writes the memo: recommendation, cases for and against, tests, kill criteria.",
      "harnessId": "anthropic-api",
      "model": "claude-opus-5",
      "dependsOn": [
        "finance",
        "customer",
        "risk"
      ],
      "timeoutMinutes": 10,
      "prompt": "Write a decision memo for the leadership team of {{vars.company}}.\n\n{{steps.frame.output}}\n\n## Financial review\n{{steps.finance.output}}\n\n## Customer review\n{{steps.customer.output}}\n\n## Execution review\n{{steps.risk.output}}\n\nSections:\n# {{issue.title}}\n## Recommendation (Go, No-go, or Go with conditions, in one line, then why in three sentences)\n## Strongest case for\n## Strongest case against\n## Assumptions to test first (with the experiment and what result would change the decision)\n## Kill criteria (the signals that should stop it after launch)\n## Open questions\n\nWhere the reviewers disagree, say so and say whose argument you weigh more and why. Under 900 words."
    }
  ]
}