diff --git a/reports/benchmark_reproducibility.json b/reports/benchmark_reproducibility.json index 8e82075..a0894bb 100644 --- a/reports/benchmark_reproducibility.json +++ b/reports/benchmark_reproducibility.json @@ -3,7 +3,7 @@ "ok": true, "generated_at": "2026-06-14", "skill_dir": ".", - "commit": "b9d502bc96a8c90de10041ca111be9094ff29a5b", + "commit": "ef0089170c1f57ab337e452a8948ceb3ae625ca1", "git_status": { "available": true, "dirty": false, @@ -17,9 +17,9 @@ "methodology_complete": true, "required_artifact_count": 24, "missing_artifact_count": 0, - "evidence_bundle_sha256": "1198b50799447f2478da882a1dfc088a0043ac699a649142f7a983e33262cfff", - "source_contract_sha256": "8a63d5ebc9ac16b7ace805fa6373f591d993a135e099eacdb209596db70f3bf2", - "archive_sha256": "501f298a9944ed9a597be439916dea3bb5812214488a46c0812bca920971f2c2", + "evidence_bundle_sha256": "f81d4dd5985eff30c70e3b4720c44a9d5b4984318463637b98645b526bfcbd6b", + "source_contract_sha256": "d3341950a7fb52a18a3a2a7453f7c697cc92e443c6604facdeeb83848d4f867c", + "archive_sha256": "c2028f0a12c88f2cae3462c80af972be934333829252faf7c92816bcc54dc541", "output_case_count": 5, "failure_disclosure_count": 3, "command_count": 21, @@ -50,7 +50,7 @@ }, "release_lock": { "ready": true, - "commit": "b9d502bc96a8c90de10041ca111be9094ff29a5b", + "commit": "ef0089170c1f57ab337e452a8948ceb3ae625ca1", "status_scope": "generation-time status before this report is written", "reason": "clean generation-time HEAD" }, @@ -60,7 +60,7 @@ "existing_count": 24, "missing_count": 0, "missing_paths": [], - "sha256": "1198b50799447f2478da882a1dfc088a0043ac699a649142f7a983e33262cfff" + "sha256": "f81d4dd5985eff30c70e3b4720c44a9d5b4984318463637b98645b526bfcbd6b" }, "methodology": { "path": "reports/benchmark_methodology.md", @@ -134,7 +134,7 @@ "path": "reports/output_execution_runs.json", "exists": true, "bytes": 7967, - "sha256": "3b833c010cc2ec6cacd2d6f59ed8d976165c6eb13124ee834e6d84793b94c87c" + "sha256": "46a65f5db66718f2f3443467c06445bee6a6269586555c307d124b03784c2d1c" }, { "label": "blind_review", @@ -169,7 +169,7 @@ "path": "reports/security_trust_report.json", "exists": true, "bytes": 98830, - "sha256": "ffa875d30f2c73a4c876aa99953b09af72c83c3e75ee04129bb05d0a3fab50ba" + "sha256": "d827b410e989366cb4e8e58257f6594feffd0f31c6e47a2a825eed73ee764ff7" }, { "label": "python_compatibility", @@ -183,14 +183,14 @@ "path": "reports/registry_audit.json", "exists": true, "bytes": 3183, - "sha256": "61fd3915f8d782ec5bda34a66bf1a67a7ef40b9a4d49534244b35f5b532fa442" + "sha256": "eba3fd128cb58de877f8e165a743192a65b5cc0c10a9086cf02fbf324a4e0fb7" }, { "label": "package_verification", "path": "reports/package_verification.json", "exists": true, "bytes": 19325, - "sha256": "9d7d30968cecf0add29f4b54e56fe642592b3fb05afe9923ccf593dde15768cb" + "sha256": "f7382b0152d969ed99fec9e0063dde02cf598f4122c4d58ec1e88198a125410b" }, { "label": "install_simulation", diff --git a/reports/benchmark_reproducibility.md b/reports/benchmark_reproducibility.md index d520977..ef2e561 100644 --- a/reports/benchmark_reproducibility.md +++ b/reports/benchmark_reproducibility.md @@ -1,9 +1,9 @@ # Benchmark Reproducibility Generated at: `2026-06-14` -Commit: `b9d502bc96a8c90de10041ca111be9094ff29a5b` +Commit: `ef0089170c1f57ab337e452a8948ceb3ae625ca1` Working tree dirty at generation: `false` -Evidence bundle SHA256: `1198b50799447f2478da882a1dfc088a0043ac699a649142f7a983e33262cfff` +Evidence bundle SHA256: `f81d4dd5985eff30c70e3b4720c44a9d5b4984318463637b98645b526bfcbd6b` ## Summary @@ -12,8 +12,8 @@ Evidence bundle SHA256: `1198b50799447f2478da882a1dfc088a0043ac699a649142f7a983e - methodology complete: `true` - required artifacts: `24` - missing artifacts: `0` -- source contract sha256: `8a63d5ebc9ac` -- archive sha256: `501f298a9944` +- source contract sha256: `d3341950a7fb` +- archive sha256: `c2028f0a12c8` - output cases: `5` - disclosed failure cases: `3` - reproduction commands: `21` @@ -48,7 +48,7 @@ This report proves local benchmark reproducibility only. It keeps external provi - algorithm: `sha256(path,label,exists,artifact_sha256)` - artifacts: `24` / `24` -- sha256: `1198b50799447f2478da882a1dfc088a0043ac699a649142f7a983e33262cfff` +- sha256: `f81d4dd5985eff30c70e3b4720c44a9d5b4984318463637b98645b526bfcbd6b` ## Methodology Sections @@ -70,15 +70,15 @@ 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 | `3b833c010cc2` | +| output_execution | `reports/output_execution_runs.json` | present | `46a65f5db667` | | 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 | `8251329e663d` | -| trust_report | `reports/security_trust_report.json` | present | `ffa875d30f2c` | +| trust_report | `reports/security_trust_report.json` | present | `d827b410e989` | | python_compatibility | `reports/python_compatibility.json` | present | `ae16e17266e4` | -| registry_audit | `reports/registry_audit.json` | present | `61fd3915f8d7` | -| package_verification | `reports/package_verification.json` | present | `9d7d30968cec` | +| registry_audit | `reports/registry_audit.json` | present | `eba3fd128cb5` | +| package_verification | `reports/package_verification.json` | present | `f7382b0152d9` | | install_simulation | `reports/install_simulation.json` | present | `8f987e805c92` | | skill_os2_audit | `reports/skill_os2_audit.json` | present | `6bb2dcb0e1e5` | | world_class_evidence_plan | `reports/world_class_evidence_plan.json` | present | `164803bb1cea` | diff --git a/reports/review-studio.json b/reports/review-studio.json index 118f108..207a0ce 100644 --- a/reports/review-studio.json +++ b/reports/review-studio.json @@ -1336,9 +1336,9 @@ "methodology_complete": true, "required_artifact_count": 24, "missing_artifact_count": 0, - "evidence_bundle_sha256": "1198b50799447f2478da882a1dfc088a0043ac699a649142f7a983e33262cfff", - "source_contract_sha256": "8a63d5ebc9ac16b7ace805fa6373f591d993a135e099eacdb209596db70f3bf2", - "archive_sha256": "501f298a9944ed9a597be439916dea3bb5812214488a46c0812bca920971f2c2", + "evidence_bundle_sha256": "f81d4dd5985eff30c70e3b4720c44a9d5b4984318463637b98645b526bfcbd6b", + "source_contract_sha256": "d3341950a7fb52a18a3a2a7453f7c697cc92e443c6604facdeeb83848d4f867c", + "archive_sha256": "c2028f0a12c88f2cae3462c80af972be934333829252faf7c92816bcc54dc541", "output_case_count": 5, "failure_disclosure_count": 3, "command_count": 21, @@ -1357,7 +1357,7 @@ "working_tree_dirty": false, "changed_file_count": 0 }, - "commit": "b9d502bc96a8c90de10041ca111be9094ff29a5b", + "commit": "ef0089170c1f57ab337e452a8948ceb3ae625ca1", "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.", @@ -4764,7 +4764,7 @@ "ok": true, "generated_at": "2026-06-14", "skill_dir": ".", - "commit": "b9d502bc96a8c90de10041ca111be9094ff29a5b", + "commit": "ef0089170c1f57ab337e452a8948ceb3ae625ca1", "git_status": { "available": true, "dirty": false, @@ -4778,9 +4778,9 @@ "methodology_complete": true, "required_artifact_count": 24, "missing_artifact_count": 0, - "evidence_bundle_sha256": "1198b50799447f2478da882a1dfc088a0043ac699a649142f7a983e33262cfff", - "source_contract_sha256": "8a63d5ebc9ac16b7ace805fa6373f591d993a135e099eacdb209596db70f3bf2", - "archive_sha256": "501f298a9944ed9a597be439916dea3bb5812214488a46c0812bca920971f2c2", + "evidence_bundle_sha256": "f81d4dd5985eff30c70e3b4720c44a9d5b4984318463637b98645b526bfcbd6b", + "source_contract_sha256": "d3341950a7fb52a18a3a2a7453f7c697cc92e443c6604facdeeb83848d4f867c", + "archive_sha256": "c2028f0a12c88f2cae3462c80af972be934333829252faf7c92816bcc54dc541", "output_case_count": 5, "failure_disclosure_count": 3, "command_count": 21, @@ -4811,7 +4811,7 @@ }, "release_lock": { "ready": true, - "commit": "b9d502bc96a8c90de10041ca111be9094ff29a5b", + "commit": "ef0089170c1f57ab337e452a8948ceb3ae625ca1", "status_scope": "generation-time status before this report is written", "reason": "clean generation-time HEAD" }, @@ -4821,7 +4821,7 @@ "existing_count": 24, "missing_count": 0, "missing_paths": [], - "sha256": "1198b50799447f2478da882a1dfc088a0043ac699a649142f7a983e33262cfff" + "sha256": "f81d4dd5985eff30c70e3b4720c44a9d5b4984318463637b98645b526bfcbd6b" }, "methodology": { "path": "reports/benchmark_methodology.md", @@ -4895,7 +4895,7 @@ "path": "reports/output_execution_runs.json", "exists": true, "bytes": 7967, - "sha256": "3b833c010cc2ec6cacd2d6f59ed8d976165c6eb13124ee834e6d84793b94c87c" + "sha256": "46a65f5db66718f2f3443467c06445bee6a6269586555c307d124b03784c2d1c" }, { "label": "blind_review", @@ -4930,7 +4930,7 @@ "path": "reports/security_trust_report.json", "exists": true, "bytes": 98830, - "sha256": "ffa875d30f2c73a4c876aa99953b09af72c83c3e75ee04129bb05d0a3fab50ba" + "sha256": "d827b410e989366cb4e8e58257f6594feffd0f31c6e47a2a825eed73ee764ff7" }, { "label": "python_compatibility", @@ -4944,14 +4944,14 @@ "path": "reports/registry_audit.json", "exists": true, "bytes": 3183, - "sha256": "61fd3915f8d782ec5bda34a66bf1a67a7ef40b9a4d49534244b35f5b532fa442" + "sha256": "eba3fd128cb58de877f8e165a743192a65b5cc0c10a9086cf02fbf324a4e0fb7" }, { "label": "package_verification", "path": "reports/package_verification.json", "exists": true, "bytes": 19325, - "sha256": "9d7d30968cecf0add29f4b54e56fe642592b3fb05afe9923ccf593dde15768cb" + "sha256": "f7382b0152d969ed99fec9e0063dde02cf598f4122c4d58ec1e88198a125410b" }, { "label": "install_simulation", diff --git a/reports/review-viewer.json b/reports/review-viewer.json index 9a66052..38f3380 100644 --- a/reports/review-viewer.json +++ b/reports/review-viewer.json @@ -923,9 +923,9 @@ "methodology_complete": true, "required_artifact_count": 24, "missing_artifact_count": 0, - "evidence_bundle_sha256": "1198b50799447f2478da882a1dfc088a0043ac699a649142f7a983e33262cfff", - "source_contract_sha256": "8a63d5ebc9ac16b7ace805fa6373f591d993a135e099eacdb209596db70f3bf2", - "archive_sha256": "501f298a9944ed9a597be439916dea3bb5812214488a46c0812bca920971f2c2", + "evidence_bundle_sha256": "f81d4dd5985eff30c70e3b4720c44a9d5b4984318463637b98645b526bfcbd6b", + "source_contract_sha256": "d3341950a7fb52a18a3a2a7453f7c697cc92e443c6604facdeeb83848d4f867c", + "archive_sha256": "c2028f0a12c88f2cae3462c80af972be934333829252faf7c92816bcc54dc541", "output_case_count": 5, "failure_disclosure_count": 3, "command_count": 21, @@ -944,7 +944,7 @@ "working_tree_dirty": false, "changed_file_count": 0 }, - "commit": "b9d502bc96a8c90de10041ca111be9094ff29a5b", + "commit": "ef0089170c1f57ab337e452a8948ceb3ae625ca1", "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 61531f7..25aebef 100644 --- a/reports/skill-overview.json +++ b/reports/skill-overview.json @@ -922,9 +922,9 @@ "methodology_complete": true, "required_artifact_count": 24, "missing_artifact_count": 0, - "evidence_bundle_sha256": "1198b50799447f2478da882a1dfc088a0043ac699a649142f7a983e33262cfff", - "source_contract_sha256": "8a63d5ebc9ac16b7ace805fa6373f591d993a135e099eacdb209596db70f3bf2", - "archive_sha256": "501f298a9944ed9a597be439916dea3bb5812214488a46c0812bca920971f2c2", + "evidence_bundle_sha256": "f81d4dd5985eff30c70e3b4720c44a9d5b4984318463637b98645b526bfcbd6b", + "source_contract_sha256": "d3341950a7fb52a18a3a2a7453f7c697cc92e443c6604facdeeb83848d4f867c", + "archive_sha256": "c2028f0a12c88f2cae3462c80af972be934333829252faf7c92816bcc54dc541", "output_case_count": 5, "failure_disclosure_count": 3, "command_count": 21, @@ -943,7 +943,7 @@ "working_tree_dirty": false, "changed_file_count": 0 }, - "commit": "b9d502bc96a8c90de10041ca111be9094ff29a5b", + "commit": "ef0089170c1f57ab337e452a8948ceb3ae625ca1", "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.",