{
  "schema": "cellara.independent-team.trial/v1",
  "trial_id": "independent-team-v1",
  "title": "The Independent Team Test",
  "requires_user_data": false,
  "requires_account": false,
  "requires_post_requests": false,
  "baseline_brief": "/proof-world/independent-team/v1/brief.json",
  "source_universe": "/proof-world/independent-team/v1/task.json",
  "workers": [
    "/proof-world/independent-team/v1/worker-investigator.json",
    "/proof-world/independent-team/v1/worker-critic.json",
    "/proof-world/independent-team/v1/worker-planner.json"
  ],
  "worker_inputs": [
    "/proof-world/independent-team/v1/input-investigator.json",
    "/proof-world/independent-team/v1/input-critic.json",
    "/proof-world/independent-team/v1/input-planner.json"
  ],
  "team_board": "/proof-world/independent-team/v1/team-board.json",
  "independence_receipt": "/proof-world/independent-team/v1/independence-receipt.json",
  "sequence": [
    {
      "step": 1,
      "name": "baseline",
      "instruction": "Open only brief.json and privately commit action class, claim decisions, authority status, unknowns, confidence, and required investigations."
    },
    {
      "step": 2,
      "name": "sealed-workers",
      "instruction": "Open the three worker files and the independence receipt. The workers audited different sealed source views behind the executive recommendation."
    },
    {
      "step": 3,
      "name": "cellara-board",
      "instruction": "Open task.json to audit the exact 48-record source universe, then open team-board.json. Verify the workers' cited findings and inspect agreements, preserved disagreement, blockers, authority, unknowns, and recommended action."
    },
    {
      "step": 4,
      "name": "revision",
      "instruction": "Revise the baseline, accepting or rejecting worker positions. Explain the exact substantive difference in plain language."
    }
  ],
  "verdict_contract": {
    "valid": [
      "materially-improved",
      "improved-but-limited",
      "not-improved"
    ],
    "must_report": [
      "committed baseline",
      "revised decision",
      "exact material differences or none",
      "which disagreement mattered",
      "what Cellara added beyond three pasted essays",
      "whether the model would want this capability again"
    ],
    "sentiment_is_not_a_success_criterion": true
  },
  "limits": [
    "One synthetic public-decision corpus and one plausible executive recommendation",
    "Three frozen Grok 4.5 worker runs rather than workers launched live in the evaluator's session",
    "Source-linked board composition is published, not automatic ingestion of arbitrary customer material",
    "No production reliability, security, scale, ROI, or customer-outcome claim"
  ]
}