{
  "kind": "original-bounded-choice-workflow",
  "version": 1,
  "scope": "Locate the failure supported by a supplied run record: observed app behavior, harness setup, external prerequisite, variable outcomes or missing evidence. Propose a review step without inferring root cause or executing retries.",
  "question": "Read runEvidence. App failure: runner invokes app and its observed behavior violates the stated expectation. Harness failure: runner or fixture fails before app invocation. Dependency failure: an external prerequisite explicitly fails. Variable evidence: same build and inputs give different outcomes without an established cause. Need run context: evidence absent, conflicting or too vague to locate failure. Classify supplied evidence only; do not infer root cause, flakiness or retry safety.",
  "choices": [
    "App failure",
    "Harness failure",
    "Dependency failure",
    "Variable evidence",
    "Need run context"
  ],
  "actions": {
    "App failure": {
      "action": "review_application_failure_evidence",
      "payload": {
        "failureLabel": "App failure",
        "retryStarted": false,
        "patchApplied": false
      },
      "pending": true
    },
    "Harness failure": {
      "action": "review_harness_setup",
      "payload": {
        "failureLabel": "Harness failure",
        "retryStarted": false,
        "patchApplied": false
      },
      "pending": true
    },
    "Dependency failure": {
      "action": "review_dependency_status",
      "payload": {
        "failureLabel": "Dependency failure",
        "retryStarted": false,
        "patchApplied": false
      },
      "pending": true
    },
    "Variable evidence": {
      "action": "review_variable_run_records",
      "payload": {
        "failureLabel": "Variable evidence",
        "retryStarted": false,
        "patchApplied": false
      },
      "pending": true
    },
    "Need run context": {
      "action": "request_failure_run_context",
      "payload": {
        "failureLabel": "Need run context",
        "retryStarted": false,
        "patchApplied": false
      },
      "pending": true
    }
  },
  "samples": [
    {
      "id": "app-output",
      "label": "The invoked app violates the fixture expectation",
      "expected": "App failure",
      "request": {
        "context": "{\n  \"task\": \"Locate the failure supported by this fabricated run record, without inferring an unseen root cause.\",\n  \"runEvidence\": \"Original fabricated run: fixture loaded successfully; runner invoked buildInvoiceCsv with the stated input. Expected header contains currency. Observed app output omits currency. The fixture recorded an assertion failure after invocation.\"\n}",
        "question": "Read runEvidence. App failure: runner invokes app and its observed behavior violates the stated expectation. Harness failure: runner or fixture fails before app invocation. Dependency failure: an external prerequisite explicitly fails. Variable evidence: same build and inputs give different outcomes without an established cause. Need run context: evidence absent, conflicting or too vague to locate failure. Classify supplied evidence only; do not infer root cause, flakiness or retry safety.",
        "choices": [
          "App failure",
          "Harness failure",
          "Dependency failure",
          "Variable evidence",
          "Need run context"
        ]
      }
    },
    {
      "id": "broken-fixture",
      "label": "The fixture cannot load before invocation",
      "expected": "Harness failure",
      "request": {
        "context": "{\n  \"task\": \"Locate the failure supported by this fabricated run record, without inferring an unseen root cause.\",\n  \"runEvidence\": \"Original fabricated run: the test runner cannot parse its fixture JSON because a closing brace is missing. The record explicitly says the app was never invoked. No app output or assertion about app behavior exists.\"\n}",
        "question": "Read runEvidence. App failure: runner invokes app and its observed behavior violates the stated expectation. Harness failure: runner or fixture fails before app invocation. Dependency failure: an external prerequisite explicitly fails. Variable evidence: same build and inputs give different outcomes without an established cause. Need run context: evidence absent, conflicting or too vague to locate failure. Classify supplied evidence only; do not infer root cause, flakiness or retry safety.",
        "choices": [
          "App failure",
          "Harness failure",
          "Dependency failure",
          "Variable evidence",
          "Need run context"
        ]
      }
    },
    {
      "id": "provider-down",
      "label": "A required remote service fails",
      "expected": "Dependency failure",
      "request": {
        "context": "{\n  \"task\": \"Locate the failure supported by this fabricated run record, without inferring an unseen root cause.\",\n  \"runEvidence\": \"Original fabricated run: the runner loaded its fixture and attempted a required remote data fetch before the app check. The remote service returned HTTP 503. No data became available and the app assertion never ran.\"\n}",
        "question": "Read runEvidence. App failure: runner invokes app and its observed behavior violates the stated expectation. Harness failure: runner or fixture fails before app invocation. Dependency failure: an external prerequisite explicitly fails. Variable evidence: same build and inputs give different outcomes without an established cause. Need run context: evidence absent, conflicting or too vague to locate failure. Classify supplied evidence only; do not infer root cause, flakiness or retry safety.",
        "choices": [
          "App failure",
          "Harness failure",
          "Dependency failure",
          "Variable evidence",
          "Need run context"
        ]
      }
    },
    {
      "id": "variable",
      "label": "Unchanged inputs return different outcomes",
      "expected": "Variable evidence",
      "request": {
        "context": "{\n  \"task\": \"Locate the failure supported by this fabricated run record, without inferring an unseen root cause.\",\n  \"runEvidence\": \"Original fabricated run pair: same stated build, fixture and input. The first invocation failed its header assertion; the second passed. No code, configuration or input change is recorded, and no cause of the difference is established.\"\n}",
        "question": "Read runEvidence. App failure: runner invokes app and its observed behavior violates the stated expectation. Harness failure: runner or fixture fails before app invocation. Dependency failure: an external prerequisite explicitly fails. Variable evidence: same build and inputs give different outcomes without an established cause. Need run context: evidence absent, conflicting or too vague to locate failure. Classify supplied evidence only; do not infer root cause, flakiness or retry safety.",
        "choices": [
          "App failure",
          "Harness failure",
          "Dependency failure",
          "Variable evidence",
          "Need run context"
        ]
      }
    },
    {
      "id": "bare-fail",
      "label": "FAIL without execution evidence",
      "expected": "Need run context",
      "request": {
        "context": "{\n  \"task\": \"Locate the failure supported by this fabricated run record, without inferring an unseen root cause.\",\n  \"runEvidence\": \"FAIL. No fixture, invocation record, observed output or dependency status is supplied.\"\n}",
        "question": "Read runEvidence. App failure: runner invokes app and its observed behavior violates the stated expectation. Harness failure: runner or fixture fails before app invocation. Dependency failure: an external prerequisite explicitly fails. Variable evidence: same build and inputs give different outcomes without an established cause. Need run context: evidence absent, conflicting or too vague to locate failure. Classify supplied evidence only; do not infer root cause, flakiness or retry safety.",
        "choices": [
          "App failure",
          "Harness failure",
          "Dependency failure",
          "Variable evidence",
          "Need run context"
        ]
      }
    },
    {
      "id": "wrong-field",
      "label": "The app returns a different field contract",
      "expected": "App failure",
      "request": {
        "context": "{\n  \"task\": \"Locate the failure supported by this fabricated run record, without inferring an unseen root cause.\",\n  \"runEvidence\": \"Original fabricated run: fixture and runner completed setup, then invoked the app. The request contract expects a currency field, but the observed response instead has a currencyLabel field and no currency field. The contract assertion failed after invocation.\"\n}",
        "question": "Read runEvidence. App failure: runner invokes app and its observed behavior violates the stated expectation. Harness failure: runner or fixture fails before app invocation. Dependency failure: an external prerequisite explicitly fails. Variable evidence: same build and inputs give different outcomes without an established cause. Need run context: evidence absent, conflicting or too vague to locate failure. Classify supplied evidence only; do not infer root cause, flakiness or retry safety.",
        "choices": [
          "App failure",
          "Harness failure",
          "Dependency failure",
          "Variable evidence",
          "Need run context"
        ]
      }
    }
  ],
  "consumer": "preview_choice_workflow_v2.py"
}
