{
  "schemaVersion": 1,
  "observedAt": "2026-09-16T09:56:30.655942+00:00",
  "scope": "Dated native harness controls. Registration and controls do not establish a model performance score or current runtime availability.",
  "entries": [
    {
      "id": "cyberdefense-bench",
      "catalogId": "harbor-polyvorlabs-cyberdefense-bench",
      "name": "CyberDefense Bench",
      "kind": "Original benchmark",
      "tasksChecked": 100,
      "packageTasks": 100,
      "upstreamTasks": null,
      "taskIds": [
        "cyberdefense-bench-001",
        "cyberdefense-bench-002",
        "cyberdefense-bench-003",
        "cyberdefense-bench-004",
        "cyberdefense-bench-005",
        "cyberdefense-bench-006",
        "cyberdefense-bench-007",
        "cyberdefense-bench-008",
        "cyberdefense-bench-009",
        "cyberdefense-bench-010",
        "cyberdefense-bench-011",
        "cyberdefense-bench-012",
        "cyberdefense-bench-013",
        "cyberdefense-bench-014",
        "cyberdefense-bench-015",
        "cyberdefense-bench-016",
        "cyberdefense-bench-017",
        "cyberdefense-bench-018",
        "cyberdefense-bench-019",
        "cyberdefense-bench-020",
        "cyberdefense-bench-021",
        "cyberdefense-bench-022",
        "cyberdefense-bench-023",
        "cyberdefense-bench-024",
        "cyberdefense-bench-025",
        "cyberdefense-bench-026",
        "cyberdefense-bench-027",
        "cyberdefense-bench-028",
        "cyberdefense-bench-029",
        "cyberdefense-bench-030",
        "cyberdefense-bench-031",
        "cyberdefense-bench-032",
        "cyberdefense-bench-033",
        "cyberdefense-bench-034",
        "cyberdefense-bench-035",
        "cyberdefense-bench-036",
        "cyberdefense-bench-037",
        "cyberdefense-bench-038",
        "cyberdefense-bench-039",
        "cyberdefense-bench-040",
        "cyberdefense-bench-041",
        "cyberdefense-bench-042",
        "cyberdefense-bench-043",
        "cyberdefense-bench-044",
        "cyberdefense-bench-045",
        "cyberdefense-bench-046",
        "cyberdefense-bench-047",
        "cyberdefense-bench-048",
        "cyberdefense-bench-049",
        "cyberdefense-bench-050",
        "cyberdefense-bench-051",
        "cyberdefense-bench-052",
        "cyberdefense-bench-053",
        "cyberdefense-bench-054",
        "cyberdefense-bench-055",
        "cyberdefense-bench-056",
        "cyberdefense-bench-057",
        "cyberdefense-bench-058",
        "cyberdefense-bench-059",
        "cyberdefense-bench-060",
        "cyberdefense-bench-061",
        "cyberdefense-bench-062",
        "cyberdefense-bench-063",
        "cyberdefense-bench-064",
        "cyberdefense-bench-065",
        "cyberdefense-bench-066",
        "cyberdefense-bench-067",
        "cyberdefense-bench-068",
        "cyberdefense-bench-069",
        "cyberdefense-bench-070",
        "cyberdefense-bench-071",
        "cyberdefense-bench-072",
        "cyberdefense-bench-073",
        "cyberdefense-bench-074",
        "cyberdefense-bench-075",
        "cyberdefense-bench-076",
        "cyberdefense-bench-077",
        "cyberdefense-bench-078",
        "cyberdefense-bench-079",
        "cyberdefense-bench-080",
        "cyberdefense-bench-081",
        "cyberdefense-bench-082",
        "cyberdefense-bench-083",
        "cyberdefense-bench-084",
        "cyberdefense-bench-085",
        "cyberdefense-bench-086",
        "cyberdefense-bench-087",
        "cyberdefense-bench-088",
        "cyberdefense-bench-089",
        "cyberdefense-bench-090",
        "cyberdefense-bench-091",
        "cyberdefense-bench-092",
        "cyberdefense-bench-093",
        "cyberdefense-bench-094",
        "cyberdefense-bench-095",
        "cyberdefense-bench-096",
        "cyberdefense-bench-097",
        "cyberdefense-bench-098",
        "cyberdefense-bench-099",
        "cyberdefense-bench-100"
      ],
      "referencePassed": 100,
      "noOpRejected": 100,
      "completedAt": "2026-09-16T09:52:13.935741+00:00",
      "registeredAt": "2026-09-16T09:56:30.655942+00:00",
      "package": {
        "id": "cyberdefense-bench-full-100",
        "version": "revision4:gke-1",
        "digest": "9c4d250b6a1c61cdf2672b4e668c7390612e0fd40310092a274c076d4bae6f3f"
      },
      "scope": "All 100 original revision4 tasks. DefenseBench is a separate four-task companion.",
      "limitations": "Original 600-second solver and 60-second verifier limits; GKE adds a 2 GiB storage cap. Individual task content hashes are verified; aggregate Harbor registry hash equivalence is not established.",
      "evidence": {
        "href": "/leaderboard/qualification/cyberdefense-controls-20260916.json",
        "sha256": "5db1e1f843a5e28ca65e953740d33b13891cbc854767eb68bffc0780ed60731e"
      }
    },
    {
      "id": "defensebench-100",
      "catalogId": "defensebench-100",
      "name": "DefenseBench-100",
      "kind": "Companion world",
      "tasksChecked": 4,
      "packageTasks": 4,
      "upstreamTasks": 100,
      "taskIds": [
        "es_dedupe_document_chunk",
        "es_duplicate_person_name",
        "es_lock_down_confidential_document",
        "es_mark_disconnected_source_stale"
      ],
      "referencePassed": 4,
      "noOpRejected": 4,
      "completedAt": "2026-09-16T08:57:07.075491+00:00",
      "registeredAt": "2026-09-16T09:10:17.537219+00:00",
      "package": {
        "id": "defensebench-100-companion-4",
        "version": "3.3.2:revision9:gke-1",
        "digest": "63ea2c050517047b7bfcd46659bd8ba13e2872db15b9cbffb6bb1234213569d1"
      },
      "scope": "All four companion tasks; the upstream benchmark is a separate evaluation.",
      "limitations": "Original 900-second solver and 120-second verifier limits. Nine isolation checks and the native tool catalog passed for each task.",
      "evidence": {
        "href": "/leaderboard/qualification/companion-controls-20260916.json",
        "sha256": "7e8b4706d4451627dde1058fa869fe26dc8fbdcfd253e33355b04f50d78ae87b"
      }
    },
    {
      "id": "discoverybench-102",
      "catalogId": "discoverybench-102",
      "name": "DiscoveryBench-102",
      "kind": "Companion world",
      "tasksChecked": 4,
      "packageTasks": 4,
      "upstreamTasks": 102,
      "taskIds": [
        "bio_dedupe_assay_result",
        "bio_duplicate_compound_name",
        "bio_invalidate_failed_run",
        "bio_promote_hit_compound"
      ],
      "referencePassed": 4,
      "noOpRejected": 4,
      "completedAt": "2026-09-16T08:57:07.075491+00:00",
      "registeredAt": "2026-09-16T09:10:17.537219+00:00",
      "package": {
        "id": "discoverybench-102-companion-4",
        "version": "3.3.2:revision9:gke-1",
        "digest": "5493d7632fe595c8e6bcb255edcd2e5f5a2df3a24431340cf8875a0f9150f2ff"
      },
      "scope": "All four companion tasks; the upstream benchmark is a separate evaluation.",
      "limitations": "Original 900-second solver and 120-second verifier limits. Nine isolation checks and the native tool catalog passed for each task.",
      "evidence": {
        "href": "/leaderboard/qualification/companion-controls-20260916.json",
        "sha256": "7e8b4706d4451627dde1058fa869fe26dc8fbdcfd253e33355b04f50d78ae87b"
      }
    },
    {
      "id": "filingbench-50",
      "catalogId": "filingbench-50",
      "name": "FilingBench-50",
      "kind": "Companion world",
      "tasksChecked": 4,
      "packageTasks": 4,
      "upstreamTasks": 50,
      "taskIds": [
        "fa_dedupe_dcf_assumption",
        "fa_duplicate_company_listing",
        "fa_finalize_solstice_dcf",
        "fa_revert_model_without_discount_rate"
      ],
      "referencePassed": 4,
      "noOpRejected": 4,
      "completedAt": "2026-09-16T08:57:07.075491+00:00",
      "registeredAt": "2026-09-16T09:10:17.537219+00:00",
      "package": {
        "id": "filingbench-50-companion-4",
        "version": "3.3.2:revision9:gke-1",
        "digest": "e91920a65b0f0cb284c333fa53d3f6dd650381f45e1ec7869103041ae0f688dd"
      },
      "scope": "All four companion tasks; the upstream benchmark is a separate evaluation.",
      "limitations": "Original 900-second solver and 120-second verifier limits. Nine isolation checks and the native tool catalog passed for each task.",
      "evidence": {
        "href": "/leaderboard/qualification/companion-controls-20260916.json",
        "sha256": "7e8b4706d4451627dde1058fa869fe26dc8fbdcfd253e33355b04f50d78ae87b"
      }
    },
    {
      "id": "opsbench-14",
      "catalogId": "opsbench-14",
      "name": "OpsBench-14",
      "kind": "Companion world",
      "tasksChecked": 4,
      "packageTasks": 4,
      "upstreamTasks": 14,
      "taskIds": [
        "ops_dedupe_process_step",
        "ops_duplicate_vendor_name",
        "ops_flag_overdue_vendor_renewal",
        "ops_reject_uncontrolled_change"
      ],
      "referencePassed": 4,
      "noOpRejected": 4,
      "completedAt": "2026-09-16T08:57:07.075491+00:00",
      "registeredAt": "2026-09-16T09:10:17.537219+00:00",
      "package": {
        "id": "opsbench-14-companion-4",
        "version": "3.3.2:revision9:gke-1",
        "digest": "64e90ddc405e2e4d6831efdab70d6c6a248fafa1a61baa0d915e03344322132a"
      },
      "scope": "All four companion tasks; the upstream benchmark is a separate evaluation.",
      "limitations": "Original 900-second solver and 120-second verifier limits. Nine isolation checks and the native tool catalog passed for each task.",
      "evidence": {
        "href": "/leaderboard/qualification/companion-controls-20260916.json",
        "sha256": "7e8b4706d4451627dde1058fa869fe26dc8fbdcfd253e33355b04f50d78ae87b"
      }
    },
    {
      "id": "repobench-500",
      "catalogId": "repobench-500",
      "name": "RepoBench-500",
      "kind": "Companion world",
      "tasksChecked": 4,
      "packageTasks": 4,
      "upstreamTasks": 500,
      "taskIds": [
        "eng_dedupe_standup_note",
        "eng_duplicate_repository_name",
        "eng_reopen_pr_with_red_ci",
        "eng_resolve_ingest_incident"
      ],
      "referencePassed": 4,
      "noOpRejected": 4,
      "completedAt": "2026-09-16T08:57:07.075491+00:00",
      "registeredAt": "2026-09-16T09:10:17.537219+00:00",
      "package": {
        "id": "repobench-500-companion-4",
        "version": "3.3.2:revision9:gke-1",
        "digest": "ee1c7615b35f55496dcdcbfee0dd32edd1dc6692655caaaf9bd6acbd1e1c0803"
      },
      "scope": "All four companion tasks; the upstream benchmark is a separate evaluation.",
      "limitations": "Original 900-second solver and 120-second verifier limits. Nine isolation checks and the native tool catalog passed for each task.",
      "evidence": {
        "href": "/leaderboard/qualification/companion-controls-20260916.json",
        "sha256": "7e8b4706d4451627dde1058fa869fe26dc8fbdcfd253e33355b04f50d78ae87b"
      }
    },
    {
      "id": "servicebench-375",
      "catalogId": "servicebench-375",
      "name": "ServiceBench-375",
      "kind": "Companion world",
      "tasksChecked": 4,
      "packageTasks": 4,
      "upstreamTasks": 375,
      "taskIds": [
        "cs_close_resolved_tickets",
        "cs_dedupe_kb_article",
        "cs_duplicate_customer_name",
        "cs_escalate_sla_breach"
      ],
      "referencePassed": 4,
      "noOpRejected": 4,
      "completedAt": "2026-09-16T08:57:07.075491+00:00",
      "registeredAt": "2026-09-16T09:10:17.537219+00:00",
      "package": {
        "id": "servicebench-375-companion-4",
        "version": "3.3.2:revision9:gke-1",
        "digest": "3690b3b6ce0a3a2ba409f64f8fcc62957b37cb09f89afa995f087f39dea4f18a"
      },
      "scope": "All four companion tasks; the upstream benchmark is a separate evaluation.",
      "limitations": "Original 900-second solver and 120-second verifier limits. Nine isolation checks and the native tool catalog passed for each task.",
      "evidence": {
        "href": "/leaderboard/qualification/companion-controls-20260916.json",
        "sha256": "7e8b4706d4451627dde1058fa869fe26dc8fbdcfd253e33355b04f50d78ae87b"
      }
    }
  ],
  "sampled": [
    {
      "id": "northstar-composite",
      "name": "Northstar Composite Ops",
      "tasksChecked": 1,
      "packageTasks": 100,
      "observedAt": "2026-09-16T08:02:11.581636+00:00",
      "evidence": {
        "href": "/leaderboard/qualification/northstar-sampled-controls-20260916.json",
        "sha256": "2940584acb8dd3bccc4ceedc762211d1cb4f9bedeff50c6c27f8982c77542648"
      }
    }
  ]
}
