diff --git a/reports/benchmark_reproducibility.json b/reports/benchmark_reproducibility.json index 7ae76c8d..1ba9d087 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": "d8e73d5b53af768d75c00ff491e1389eb24af8d5", + "commit": "ae5ce88ea7f4d002670ce9d3cb2bb339017902d7", "git_status": { "available": true, "dirty": false, @@ -17,9 +17,9 @@ "methodology_complete": true, "required_artifact_count": 24, "missing_artifact_count": 0, - "evidence_bundle_sha256": "667271922a07f98d6a759e3b90fe72de75256f70b5fe3a454d4e1dab3eb1fc68", + "evidence_bundle_sha256": "bab9f7cd9fc387621005ec17e41186d232af6be1fb154824493e5c5b5cac6027", "source_contract_sha256": "30502ca01a3d74140f6ad23d9d5fd499a8c80dcd4b8428ad1c9a812d2e3f9b5e", - "archive_sha256": "6c34f2cfa88c7a018cbe02e41c98b41d32d39293b1d1e6c75a7a780d8add3a6e", + "archive_sha256": "f1c499c597e88a4e1121d3464f5bcedfff2530128edf7a81f95602dda53e1874", "output_case_count": 5, "failure_disclosure_count": 3, "command_count": 22, @@ -54,7 +54,7 @@ }, "release_lock": { "ready": true, - "commit": "d8e73d5b53af768d75c00ff491e1389eb24af8d5", + "commit": "ae5ce88ea7f4d002670ce9d3cb2bb339017902d7", "status_scope": "generation-time status before this report is written", "reason": "clean generation-time HEAD" }, @@ -64,7 +64,7 @@ "existing_count": 24, "missing_count": 0, "missing_paths": [], - "sha256": "667271922a07f98d6a759e3b90fe72de75256f70b5fe3a454d4e1dab3eb1fc68" + "sha256": "bab9f7cd9fc387621005ec17e41186d232af6be1fb154824493e5c5b5cac6027" }, "methodology": { "path": "reports/benchmark_methodology.md", @@ -138,7 +138,7 @@ "path": "reports/output_execution_runs.json", "exists": true, "bytes": 7967, - "sha256": "3dd95648013bb7a04649e8c6ea05b5899a88f0883f2853a69bc7698817f08888" + "sha256": "ddb48310b50baee9dec234264e789cf75024016a6c1073ea1e391ca0b0381c53" }, { "label": "blind_review", @@ -187,14 +187,14 @@ "path": "reports/registry_audit.json", "exists": true, "bytes": 3183, - "sha256": "88c486a23471a7f81f3a00ccc2cb13c2dd19c22d29f8fc6cf0b938cc7904bb29" + "sha256": "afb07b53145703e057484fc6ca8c9efa3fb348d593d9449853d8e5daa7900f30" }, { "label": "package_verification", "path": "reports/package_verification.json", "exists": true, "bytes": 19338, - "sha256": "005aee7807584f2bb62064eaf4def5ca20cd11d51a0011c0d9a4f00e66677b80" + "sha256": "affc7184918121ab719b916c87b5c8c058902a2d4747356bf3c34a60a32e769f" }, { "label": "install_simulation", diff --git a/reports/benchmark_reproducibility.md b/reports/benchmark_reproducibility.md index cd002879..52f832ed 100644 --- a/reports/benchmark_reproducibility.md +++ b/reports/benchmark_reproducibility.md @@ -1,9 +1,9 @@ # Benchmark Reproducibility Generated at: `2026-06-16` -Commit: `d8e73d5b53af768d75c00ff491e1389eb24af8d5` +Commit: `ae5ce88ea7f4d002670ce9d3cb2bb339017902d7` Working tree dirty at generation: `false` -Evidence bundle SHA256: `667271922a07f98d6a759e3b90fe72de75256f70b5fe3a454d4e1dab3eb1fc68` +Evidence bundle SHA256: `bab9f7cd9fc387621005ec17e41186d232af6be1fb154824493e5c5b5cac6027` ## Summary @@ -13,7 +13,7 @@ Evidence bundle SHA256: `667271922a07f98d6a759e3b90fe72de75256f70b5fe3a454d4e1da - required artifacts: `24` - missing artifacts: `0` - source contract sha256: `30502ca01a3d` -- archive sha256: `6c34f2cfa88c` +- archive sha256: `f1c499c597e8` - output cases: `5` - disclosed failure cases: `3` - reproduction commands: `22` @@ -50,7 +50,7 @@ This report proves local benchmark reproducibility only. It keeps external provi - algorithm: `sha256(path,label,exists,artifact_sha256)` - artifacts: `24` / `24` -- sha256: `667271922a07f98d6a759e3b90fe72de75256f70b5fe3a454d4e1dab3eb1fc68` +- sha256: `bab9f7cd9fc387621005ec17e41186d232af6be1fb154824493e5c5b5cac6027` ## Methodology Sections @@ -72,15 +72,15 @@ 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 | `3dd95648013b` | +| output_execution | `reports/output_execution_runs.json` | present | `ddb48310b50b` | | 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 | `154c7e8534de` | | python_compatibility | `reports/python_compatibility.json` | present | `471c481ff9f9` | -| registry_audit | `reports/registry_audit.json` | present | `88c486a23471` | -| package_verification | `reports/package_verification.json` | present | `005aee780758` | +| registry_audit | `reports/registry_audit.json` | present | `afb07b531457` | +| package_verification | `reports/package_verification.json` | present | `affc71849181` | | install_simulation | `reports/install_simulation.json` | present | `28ceb014e202` | | skill_os2_audit | `reports/skill_os2_audit.json` | present | `57536bc67370` | | world_class_evidence_plan | `reports/world_class_evidence_plan.json` | present | `76a3f8e2b12b` | diff --git a/reports/evidence_consistency.json b/reports/evidence_consistency.json index 495a4b74..27faaf38 100644 --- a/reports/evidence_consistency.json +++ b/reports/evidence_consistency.json @@ -53,8 +53,8 @@ "key": "overview-benchmark-commit", "label": "overview embeds the benchmark commit", "status": "pass", - "expected": "d8e73d5b53af768d75c00ff491e1389eb24af8d5", - "actual": "d8e73d5b53af768d75c00ff491e1389eb24af8d5", + "expected": "ae5ce88ea7f4d002670ce9d3cb2bb339017902d7", + "actual": "ae5ce88ea7f4d002670ce9d3cb2bb339017902d7", "paths": [ "reports/benchmark_reproducibility.json", "reports/skill-overview.json" @@ -70,7 +70,7 @@ "required_artifact_count": 24, "missing_artifact_count": 0, "source_contract_sha256": "30502ca01a3d74140f6ad23d9d5fd499a8c80dcd4b8428ad1c9a812d2e3f9b5e", - "archive_sha256": "6c34f2cfa88c7a018cbe02e41c98b41d32d39293b1d1e6c75a7a780d8add3a6e", + "archive_sha256": "f1c499c597e88a4e1121d3464f5bcedfff2530128edf7a81f95602dda53e1874", "world_class_ledger_pending_count": 4, "world_class_source_check_count": 13, "world_class_source_pass_count": 6, @@ -83,7 +83,7 @@ "required_artifact_count": 24, "missing_artifact_count": 0, "source_contract_sha256": "30502ca01a3d74140f6ad23d9d5fd499a8c80dcd4b8428ad1c9a812d2e3f9b5e", - "archive_sha256": "6c34f2cfa88c7a018cbe02e41c98b41d32d39293b1d1e6c75a7a780d8add3a6e", + "archive_sha256": "f1c499c597e88a4e1121d3464f5bcedfff2530128edf7a81f95602dda53e1874", "world_class_ledger_pending_count": 4, "world_class_source_check_count": 13, "world_class_source_pass_count": 6, @@ -199,8 +199,8 @@ "key": "interpretation-benchmark-commit", "label": "interpretation embeds the benchmark commit", "status": "pass", - "expected": "d8e73d5b53af768d75c00ff491e1389eb24af8d5", - "actual": "d8e73d5b53af768d75c00ff491e1389eb24af8d5", + "expected": "ae5ce88ea7f4d002670ce9d3cb2bb339017902d7", + "actual": "ae5ce88ea7f4d002670ce9d3cb2bb339017902d7", "paths": [ "reports/benchmark_reproducibility.json", "reports/skill-interpretation.json" @@ -216,7 +216,7 @@ "required_artifact_count": 24, "missing_artifact_count": 0, "source_contract_sha256": "30502ca01a3d74140f6ad23d9d5fd499a8c80dcd4b8428ad1c9a812d2e3f9b5e", - "archive_sha256": "6c34f2cfa88c7a018cbe02e41c98b41d32d39293b1d1e6c75a7a780d8add3a6e", + "archive_sha256": "f1c499c597e88a4e1121d3464f5bcedfff2530128edf7a81f95602dda53e1874", "world_class_ledger_pending_count": 4, "world_class_source_check_count": 13, "world_class_source_pass_count": 6, @@ -229,7 +229,7 @@ "required_artifact_count": 24, "missing_artifact_count": 0, "source_contract_sha256": "30502ca01a3d74140f6ad23d9d5fd499a8c80dcd4b8428ad1c9a812d2e3f9b5e", - "archive_sha256": "6c34f2cfa88c7a018cbe02e41c98b41d32d39293b1d1e6c75a7a780d8add3a6e", + "archive_sha256": "f1c499c597e88a4e1121d3464f5bcedfff2530128edf7a81f95602dda53e1874", "world_class_ledger_pending_count": 4, "world_class_source_check_count": 13, "world_class_source_pass_count": 6, diff --git a/reports/review-studio.json b/reports/review-studio.json index 2e617afb..bf098877 100644 --- a/reports/review-studio.json +++ b/reports/review-studio.json @@ -1680,9 +1680,9 @@ "methodology_complete": true, "required_artifact_count": 24, "missing_artifact_count": 0, - "evidence_bundle_sha256": "667271922a07f98d6a759e3b90fe72de75256f70b5fe3a454d4e1dab3eb1fc68", + "evidence_bundle_sha256": "bab9f7cd9fc387621005ec17e41186d232af6be1fb154824493e5c5b5cac6027", "source_contract_sha256": "30502ca01a3d74140f6ad23d9d5fd499a8c80dcd4b8428ad1c9a812d2e3f9b5e", - "archive_sha256": "6c34f2cfa88c7a018cbe02e41c98b41d32d39293b1d1e6c75a7a780d8add3a6e", + "archive_sha256": "f1c499c597e88a4e1121d3464f5bcedfff2530128edf7a81f95602dda53e1874", "output_case_count": 5, "failure_disclosure_count": 3, "command_count": 22, @@ -1704,7 +1704,7 @@ "working_tree_dirty": false, "changed_file_count": 0 }, - "commit": "d8e73d5b53af768d75c00ff491e1389eb24af8d5", + "commit": "ae5ce88ea7f4d002670ce9d3cb2bb339017902d7", "missing_artifacts": [], "limitations": [ "The git commit and dirty flag are generation-time context; the evidence bundle hash is the durable artifact anchor inside a committed report.", @@ -5299,7 +5299,7 @@ "ok": true, "generated_at": "2026-06-16", "skill_dir": ".", - "commit": "d8e73d5b53af768d75c00ff491e1389eb24af8d5", + "commit": "ae5ce88ea7f4d002670ce9d3cb2bb339017902d7", "git_status": { "available": true, "dirty": false, @@ -5313,9 +5313,9 @@ "methodology_complete": true, "required_artifact_count": 24, "missing_artifact_count": 0, - "evidence_bundle_sha256": "667271922a07f98d6a759e3b90fe72de75256f70b5fe3a454d4e1dab3eb1fc68", + "evidence_bundle_sha256": "bab9f7cd9fc387621005ec17e41186d232af6be1fb154824493e5c5b5cac6027", "source_contract_sha256": "30502ca01a3d74140f6ad23d9d5fd499a8c80dcd4b8428ad1c9a812d2e3f9b5e", - "archive_sha256": "6c34f2cfa88c7a018cbe02e41c98b41d32d39293b1d1e6c75a7a780d8add3a6e", + "archive_sha256": "f1c499c597e88a4e1121d3464f5bcedfff2530128edf7a81f95602dda53e1874", "output_case_count": 5, "failure_disclosure_count": 3, "command_count": 22, @@ -5350,7 +5350,7 @@ }, "release_lock": { "ready": true, - "commit": "d8e73d5b53af768d75c00ff491e1389eb24af8d5", + "commit": "ae5ce88ea7f4d002670ce9d3cb2bb339017902d7", "status_scope": "generation-time status before this report is written", "reason": "clean generation-time HEAD" }, @@ -5360,7 +5360,7 @@ "existing_count": 24, "missing_count": 0, "missing_paths": [], - "sha256": "667271922a07f98d6a759e3b90fe72de75256f70b5fe3a454d4e1dab3eb1fc68" + "sha256": "bab9f7cd9fc387621005ec17e41186d232af6be1fb154824493e5c5b5cac6027" }, "methodology": { "path": "reports/benchmark_methodology.md", @@ -5434,7 +5434,7 @@ "path": "reports/output_execution_runs.json", "exists": true, "bytes": 7967, - "sha256": "3dd95648013bb7a04649e8c6ea05b5899a88f0883f2853a69bc7698817f08888" + "sha256": "ddb48310b50baee9dec234264e789cf75024016a6c1073ea1e391ca0b0381c53" }, { "label": "blind_review", @@ -5483,14 +5483,14 @@ "path": "reports/registry_audit.json", "exists": true, "bytes": 3183, - "sha256": "88c486a23471a7f81f3a00ccc2cb13c2dd19c22d29f8fc6cf0b938cc7904bb29" + "sha256": "afb07b53145703e057484fc6ca8c9efa3fb348d593d9449853d8e5daa7900f30" }, { "label": "package_verification", "path": "reports/package_verification.json", "exists": true, "bytes": 19338, - "sha256": "005aee7807584f2bb62064eaf4def5ca20cd11d51a0011c0d9a4f00e66677b80" + "sha256": "affc7184918121ab719b916c87b5c8c058902a2d4747356bf3c34a60a32e769f" }, { "label": "install_simulation", diff --git a/reports/review-viewer.json b/reports/review-viewer.json index d20f30eb..2b2926c8 100644 --- a/reports/review-viewer.json +++ b/reports/review-viewer.json @@ -1000,9 +1000,9 @@ "methodology_complete": true, "required_artifact_count": 24, "missing_artifact_count": 0, - "evidence_bundle_sha256": "667271922a07f98d6a759e3b90fe72de75256f70b5fe3a454d4e1dab3eb1fc68", + "evidence_bundle_sha256": "bab9f7cd9fc387621005ec17e41186d232af6be1fb154824493e5c5b5cac6027", "source_contract_sha256": "30502ca01a3d74140f6ad23d9d5fd499a8c80dcd4b8428ad1c9a812d2e3f9b5e", - "archive_sha256": "6c34f2cfa88c7a018cbe02e41c98b41d32d39293b1d1e6c75a7a780d8add3a6e", + "archive_sha256": "f1c499c597e88a4e1121d3464f5bcedfff2530128edf7a81f95602dda53e1874", "output_case_count": 5, "failure_disclosure_count": 3, "command_count": 22, @@ -1024,7 +1024,7 @@ "working_tree_dirty": false, "changed_file_count": 0 }, - "commit": "d8e73d5b53af768d75c00ff491e1389eb24af8d5", + "commit": "ae5ce88ea7f4d002670ce9d3cb2bb339017902d7", "missing_artifacts": [], "limitations": [ "The git commit and dirty flag are generation-time context; the evidence bundle hash is the durable artifact anchor inside a committed report.", diff --git a/reports/skill-interpretation.json b/reports/skill-interpretation.json index 0beb8d54..e49f5651 100644 --- a/reports/skill-interpretation.json +++ b/reports/skill-interpretation.json @@ -1004,9 +1004,9 @@ "methodology_complete": true, "required_artifact_count": 24, "missing_artifact_count": 0, - "evidence_bundle_sha256": "667271922a07f98d6a759e3b90fe72de75256f70b5fe3a454d4e1dab3eb1fc68", + "evidence_bundle_sha256": "bab9f7cd9fc387621005ec17e41186d232af6be1fb154824493e5c5b5cac6027", "source_contract_sha256": "30502ca01a3d74140f6ad23d9d5fd499a8c80dcd4b8428ad1c9a812d2e3f9b5e", - "archive_sha256": "6c34f2cfa88c7a018cbe02e41c98b41d32d39293b1d1e6c75a7a780d8add3a6e", + "archive_sha256": "f1c499c597e88a4e1121d3464f5bcedfff2530128edf7a81f95602dda53e1874", "output_case_count": 5, "failure_disclosure_count": 3, "command_count": 22, @@ -1028,7 +1028,7 @@ "working_tree_dirty": false, "changed_file_count": 0 }, - "commit": "d8e73d5b53af768d75c00ff491e1389eb24af8d5", + "commit": "ae5ce88ea7f4d002670ce9d3cb2bb339017902d7", "missing_artifacts": [], "limitations": [ "The git commit and dirty flag are generation-time context; the evidence bundle hash is the durable artifact anchor inside a committed report.", diff --git a/reports/skill-overview.json b/reports/skill-overview.json index 81f7b7df..8166e342 100644 --- a/reports/skill-overview.json +++ b/reports/skill-overview.json @@ -999,9 +999,9 @@ "methodology_complete": true, "required_artifact_count": 24, "missing_artifact_count": 0, - "evidence_bundle_sha256": "667271922a07f98d6a759e3b90fe72de75256f70b5fe3a454d4e1dab3eb1fc68", + "evidence_bundle_sha256": "bab9f7cd9fc387621005ec17e41186d232af6be1fb154824493e5c5b5cac6027", "source_contract_sha256": "30502ca01a3d74140f6ad23d9d5fd499a8c80dcd4b8428ad1c9a812d2e3f9b5e", - "archive_sha256": "6c34f2cfa88c7a018cbe02e41c98b41d32d39293b1d1e6c75a7a780d8add3a6e", + "archive_sha256": "f1c499c597e88a4e1121d3464f5bcedfff2530128edf7a81f95602dda53e1874", "output_case_count": 5, "failure_disclosure_count": 3, "command_count": 22, @@ -1023,7 +1023,7 @@ "working_tree_dirty": false, "changed_file_count": 0 }, - "commit": "d8e73d5b53af768d75c00ff491e1389eb24af8d5", + "commit": "ae5ce88ea7f4d002670ce9d3cb2bb339017902d7", "missing_artifacts": [], "limitations": [ "The git commit and dirty flag are generation-time context; the evidence bundle hash is the durable artifact anchor inside a committed report.",