diff --git a/reports/benchmark_reproducibility.json b/reports/benchmark_reproducibility.json index 05b662a..8c733c1 100644 --- a/reports/benchmark_reproducibility.json +++ b/reports/benchmark_reproducibility.json @@ -3,7 +3,7 @@ "ok": true, "generated_at": "2026-06-16", "skill_dir": ".", - "commit": "904969b91f6ce1407ea5d973f359b73d5f80ca7a", + "commit": "aee549103223c157e934d31f9c4dc5b5effb0050", "git_status": { "available": true, "dirty": false, @@ -17,9 +17,9 @@ "methodology_complete": true, "required_artifact_count": 25, "missing_artifact_count": 0, - "evidence_bundle_sha256": "8eaf9f0bedd393943fce51d00f80191be7cb6a92b62aab3d6e7f341bd6bdd31c", - "source_contract_sha256": "7de584946cac8f2ed241aa331ba4c0c2d69fa1c4eae2c06b92882b8af6ad07d7", - "archive_sha256": "1d4626e5f8a27d7c708bb0e13f6f1aab44db25f9c5154b548fc5e70a52042a35", + "evidence_bundle_sha256": "2daca5a423774255219e5f5fbfcbcd6a2fea7d1e36c567c4ca1bcf639b51df23", + "source_contract_sha256": "c29e7a583ec56e9ab8ec665f9a4e152e4135f0babb4806b4564e6aface56dada", + "archive_sha256": "ff4e46e4a22fee8a7bcd7c155b00df4ad4df7d9a85c939d657027fa6dea59d6d", "output_case_count": 5, "failure_disclosure_count": 3, "command_count": 23, @@ -54,7 +54,7 @@ }, "release_lock": { "ready": true, - "commit": "904969b91f6ce1407ea5d973f359b73d5f80ca7a", + "commit": "aee549103223c157e934d31f9c4dc5b5effb0050", "status_scope": "generation-time status before this report is written", "reason": "clean generation-time HEAD" }, @@ -64,7 +64,7 @@ "existing_count": 25, "missing_count": 0, "missing_paths": [], - "sha256": "8eaf9f0bedd393943fce51d00f80191be7cb6a92b62aab3d6e7f341bd6bdd31c" + "sha256": "2daca5a423774255219e5f5fbfcbcd6a2fea7d1e36c567c4ca1bcf639b51df23" }, "methodology": { "path": "reports/benchmark_methodology.md", @@ -138,7 +138,7 @@ "path": "reports/output_execution_runs.json", "exists": true, "bytes": 7967, - "sha256": "c1f94258e640e449eca383be009ff54d883f44bace9738d7717392043b1d3e71" + "sha256": "b09234cddcbdb007a75424e934aebb9154d523552f9a2cc4398d6856bd9b1265" }, { "label": "blind_review", @@ -172,43 +172,43 @@ "label": "trust_report", "path": "reports/security_trust_report.json", "exists": true, - "bytes": 113805, - "sha256": "47c789ac5ff9404eaeb16fa762a805ea8d12c923dd54c74f97d1f3ccb312a556" + "bytes": 114928, + "sha256": "14b3fdf245332932bedc6aa603c8019be67e0c2b64c18a53321b2aa79eac3efb" }, { "label": "python_compatibility", "path": "reports/python_compatibility.json", "exists": true, - "bytes": 24156, - "sha256": "9392cf6f3c9042044ddd5e2467e4e81f7d9e31587265d906359d4ca80c9e0958" + "bytes": 24532, + "sha256": "2740264f51ef8b1b1d197c7a03cc78095b04b846b0f1909266f12a14db323552" }, { "label": "registry_audit", "path": "reports/registry_audit.json", "exists": true, "bytes": 3183, - "sha256": "462d39d38d11e4f885ba965a320a61dbf0d4ad9b8bd6fe909ee0c039b3058f08" + "sha256": "70e1c0be9834f3447d8ab6131e5e816b6a5d1710fa59caae1441077c8468545f" }, { "label": "package_verification", "path": "reports/package_verification.json", "exists": true, "bytes": 19338, - "sha256": "bd6b8467def56d3667aa931c5ace75227dc9f4345e0ab02632d1d99e5aeac91a" + "sha256": "713523fd3605a717084f8ec3e2041e2c58fe46eb4d34b605458128792fc183b6" }, { "label": "install_simulation", "path": "reports/install_simulation.json", "exists": true, "bytes": 8604, - "sha256": "1ad04c50c23ff7c76115b7218d22c4c86ea10bf2bcefb34d4b2d6d10122f350b" + "sha256": "2556db81df85b8c7b8a2df34a8f116459cf0f6aad3c47c4f0f2c7202f8065be1" }, { "label": "skill_os2_audit", "path": "reports/skill_os2_audit.json", "exists": true, "bytes": 14310, - "sha256": "e7d9819e12a79f97a0e708f6401db7ede878deff3a9b2aa3fe2f5aba09fdb15e" + "sha256": "15fd40e0809088f83fd80797864ec256d6860007fd46cf2507475c4a72da70e8" }, { "label": "world_class_evidence_plan", @@ -270,8 +270,8 @@ "label": "world_class_claim_guard", "path": "reports/world_class_claim_guard.json", "exists": true, - "bytes": 18218, - "sha256": "7359ad2ef09dc0863014aab541a92a1f3e55f18f0c0446816c1c7333cba256e9" + "bytes": 18406, + "sha256": "704aba252d55e9b9f546392ce7f51b91f3b74d3384501d7f80a74e14c5b50acd" } ], "missing_artifacts": [], diff --git a/reports/benchmark_reproducibility.md b/reports/benchmark_reproducibility.md index 44432ed..b034bff 100644 --- a/reports/benchmark_reproducibility.md +++ b/reports/benchmark_reproducibility.md @@ -1,9 +1,9 @@ # Benchmark Reproducibility Generated at: `2026-06-16` -Commit: `904969b91f6ce1407ea5d973f359b73d5f80ca7a` +Commit: `aee549103223c157e934d31f9c4dc5b5effb0050` Working tree dirty at generation: `false` -Evidence bundle SHA256: `8eaf9f0bedd393943fce51d00f80191be7cb6a92b62aab3d6e7f341bd6bdd31c` +Evidence bundle SHA256: `2daca5a423774255219e5f5fbfcbcd6a2fea7d1e36c567c4ca1bcf639b51df23` ## Summary @@ -12,8 +12,8 @@ Evidence bundle SHA256: `8eaf9f0bedd393943fce51d00f80191be7cb6a92b62aab3d6e7f341 - methodology complete: `true` - required artifacts: `25` - missing artifacts: `0` -- source contract sha256: `7de584946cac` -- archive sha256: `1d4626e5f8a2` +- source contract sha256: `c29e7a583ec5` +- archive sha256: `ff4e46e4a22f` - output cases: `5` - disclosed failure cases: `3` - reproduction commands: `23` @@ -50,7 +50,7 @@ This report proves local benchmark reproducibility only. It keeps external provi - algorithm: `sha256(path,label,exists,artifact_sha256)` - artifacts: `25` / `25` -- sha256: `8eaf9f0bedd393943fce51d00f80191be7cb6a92b62aab3d6e7f341bd6bdd31c` +- sha256: `2daca5a423774255219e5f5fbfcbcd6a2fea7d1e36c567c4ca1bcf639b51df23` ## Methodology Sections @@ -72,17 +72,17 @@ This report proves local benchmark reproducibility only. It keeps external provi | output_cases | `evals/output/cases.jsonl` | present | `a6ae96857116` | | output_schema | `evals/output/schema.json` | present | `8ee340c95064` | | output_scorecard | `reports/output_quality_scorecard.json` | present | `0806258a8e08` | -| output_execution | `reports/output_execution_runs.json` | present | `c1f94258e640` | +| output_execution | `reports/output_execution_runs.json` | present | `b09234cddcbd` | | blind_review | `reports/output_blind_review_pack.json` | present | `bbe2db8ec277` | | review_adjudication | `reports/output_review_adjudication.json` | present | `240485a721af` | | trigger_scorecard | `reports/route_scorecard.json` | present | `c164e83e36d0` | | runtime_conformance | `reports/conformance_matrix.json` | present | `97f9ba949c23` | -| trust_report | `reports/security_trust_report.json` | present | `47c789ac5ff9` | -| python_compatibility | `reports/python_compatibility.json` | present | `9392cf6f3c90` | -| registry_audit | `reports/registry_audit.json` | present | `462d39d38d11` | -| package_verification | `reports/package_verification.json` | present | `bd6b8467def5` | -| install_simulation | `reports/install_simulation.json` | present | `1ad04c50c23f` | -| skill_os2_audit | `reports/skill_os2_audit.json` | present | `e7d9819e12a7` | +| trust_report | `reports/security_trust_report.json` | present | `14b3fdf24533` | +| python_compatibility | `reports/python_compatibility.json` | present | `2740264f51ef` | +| registry_audit | `reports/registry_audit.json` | present | `70e1c0be9834` | +| package_verification | `reports/package_verification.json` | present | `713523fd3605` | +| install_simulation | `reports/install_simulation.json` | present | `2556db81df85` | +| skill_os2_audit | `reports/skill_os2_audit.json` | present | `15fd40e08090` | | world_class_evidence_plan | `reports/world_class_evidence_plan.json` | present | `76a3f8e2b12b` | | world_class_evidence_ledger | `reports/world_class_evidence_ledger.json` | present | `1cb74eaa978b` | | world_class_evidence_intake | `reports/world_class_evidence_intake.json` | present | `99a0ba234d59` | @@ -91,7 +91,7 @@ This report proves local benchmark reproducibility only. It keeps external provi | world_class_operator_runbook | `reports/world_class_operator_runbook.json` | present | `f5ca32d5ea8b` | | world_class_operator_runbook_markdown | `reports/world_class_operator_runbook.md` | present | `260d5e03c52e` | | world_class_operator_runbook_html | `reports/world_class_operator_runbook.html` | present | `04cc091b113f` | -| world_class_claim_guard | `reports/world_class_claim_guard.json` | present | `7359ad2ef09d` | +| world_class_claim_guard | `reports/world_class_claim_guard.json` | present | `704aba252d55` | ## Reproduction Commands